From ced918eb748ce30b3aace549fd17540e40ffdca0 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 17 Feb 2010 16:47:10 +0000 Subject: i8253: Convert i8253_lock to raw_spinlock i8253_lock needs to be a real spinlock in preempt-rt, i.e. it can not be converted to a sleeping lock. Convert it to raw_spinlock and fix up all users. Signed-off-by: Thomas Gleixner Acked-by: Ralf Baechle Acked-by: Dmitry Torokhov Acked-by: Takashi Iwai Cc: Jens Axboe LKML-Reference: <20100217163751.030764372@linutronix.de> --- drivers/block/hd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/hd.c b/drivers/block/hd.c index 5116c65..b9868ad 100644 --- a/drivers/block/hd.c +++ b/drivers/block/hd.c @@ -165,12 +165,12 @@ unsigned long read_timer(void) unsigned long t, flags; int i; - spin_lock_irqsave(&i8253_lock, flags); + raw_spin_lock_irqsave(&i8253_lock, flags); t = jiffies * 11932; outb_p(0, 0x43); i = inb_p(0x40); i |= inb(0x40) << 8; - spin_unlock_irqrestore(&i8253_lock, flags); + raw_spin_unlock_irqrestore(&i8253_lock, flags); return(t - i); } #endif -- cgit v1.1 From 7ac314c82f552eefebaa91c9fffe8c0d435641b9 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Thu, 22 Apr 2010 14:27:23 +0200 Subject: drbd: fix memory leak We leak memory if "--dry-run" is not supported by the peer. Signed-off-by: Dan Carpenter Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_main.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 67e0fc5..93d1f9b 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1695,6 +1695,7 @@ int drbd_send_protocol(struct drbd_conf *mdev) cf |= CF_DRY_RUN; else { dev_err(DEV, "--dry-run is not supported by peer"); + kfree(p); return 0; } } -- cgit v1.1 From 7e2455c1a123ceadbb35150a610d61e8443fd340 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 22 Apr 2010 14:50:23 +0200 Subject: drbd: Terminate a connection early if sending the protocol fails Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_receiver.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index ed9f1de..3f096e7 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -899,7 +899,8 @@ retry: drbd_thread_start(&mdev->asender); - drbd_send_protocol(mdev); + if (!drbd_send_protocol(mdev)) + return -1; drbd_send_sync_param(mdev, &mdev->sync_conf); drbd_send_sizes(mdev, 0); drbd_send_uuids(mdev); -- cgit v1.1 From fbd9b09a177a481eda256447c881f014f29034fe Mon Sep 17 00:00:00 2001 From: Dmitry Monakhov Date: Wed, 28 Apr 2010 17:55:06 +0400 Subject: blkdev: generalize flags for blkdev_issue_fn functions The patch just convert all blkdev_issue_xxx function to common set of flags. Wait/allocation semantics preserved. Signed-off-by: Dmitry Monakhov Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_int.h | 3 ++- drivers/block/drbd/drbd_receiver.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index e5e86a7..d6f1ae3 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -2251,7 +2251,8 @@ static inline void drbd_md_flush(struct drbd_conf *mdev) if (test_bit(MD_NO_BARRIER, &mdev->flags)) return; - r = blkdev_issue_flush(mdev->ldev->md_bdev, NULL); + r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL, + BLKDEV_IFL_WAIT); if (r) { set_bit(MD_NO_BARRIER, &mdev->flags); dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r); diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index ed9f1de..54f56ea 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -945,7 +945,8 @@ static enum finish_epoch drbd_flush_after_epoch(struct drbd_conf *mdev, struct d int rv; if (mdev->write_ordering >= WO_bdev_flush && get_ldev(mdev)) { - rv = blkdev_issue_flush(mdev->ldev->backing_bdev, NULL); + rv = blkdev_issue_flush(mdev->ldev->backing_bdev, GFP_KERNEL, + NULL, BLKDEV_IFL_WAIT); if (rv) { dev_err(DEV, "local disk flush failed with status %d\n", rv); /* would rather check on EOPNOTSUPP, but that is not reliable. -- cgit v1.1 From f80a0ca6ad8f2800453e819dafa09a0ed9e56850 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Wed, 28 Apr 2010 14:36:41 +0200 Subject: pktcdvd: improve BKL and compat_ioctl.c usage The pktcdvd driver uses proper locking and does not need the BKL in the ioctl and llseek functions of the character device, so kill both. Moving the compat_ioctl handling from common code into the driver itself fixes build problems when CONFIG_BLOCK is disabled. Acked-by: Randy Dunlap Signed-off-by: Arnd Bergmann Signed-off-by: Linus Torvalds --- drivers/block/pktcdvd.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c index ddf1942..8a549db 100644 --- a/drivers/block/pktcdvd.c +++ b/drivers/block/pktcdvd.c @@ -48,6 +48,7 @@ #include #include #include +#include #include #include #include @@ -2984,7 +2985,7 @@ static void pkt_get_status(struct pkt_ctrl_command *ctrl_cmd) mutex_unlock(&ctl_mutex); } -static int pkt_ctl_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg) +static long pkt_ctl_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { void __user *argp = (void __user *)arg; struct pkt_ctrl_command ctrl_cmd; @@ -3021,10 +3022,20 @@ static int pkt_ctl_ioctl(struct inode *inode, struct file *file, unsigned int cm return ret; } +#ifdef CONFIG_COMPAT +static long pkt_ctl_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) +{ + return pkt_ctl_ioctl(file, cmd, (unsigned long)compat_ptr(arg)); +} +#endif static const struct file_operations pkt_ctl_fops = { - .ioctl = pkt_ctl_ioctl, - .owner = THIS_MODULE, + .open = nonseekable_open, + .unlocked_ioctl = pkt_ctl_ioctl, +#ifdef CONFIG_COMPAT + .compat_ioctl = pkt_ctl_compat_ioctl, +#endif + .owner = THIS_MODULE, }; static struct miscdevice pkt_misc = { -- cgit v1.1 From 5c3c7e64bbcb60db39d0d55c8fab39ef41d41c30 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Sat, 10 Apr 2010 02:10:09 +0200 Subject: drbd: don't expose failed local READ to upper layers fix regression introduced in 8.3.3: commit a9b17323f2875f5d9b132c2b476a750bf44b10c7 Author: Lars Ellenberg Date: Wed Aug 12 15:18:33 2009 +0200 out-of-spinlock completion of master bio : (bio_rw(bio) == READA) ? read_completed_with_error : read_ahead_completed_with_error; is obviously not what was intended. No one noticed because of * page-cache at work, * local RAIDs Impact: Failed local READs are not retried remotely, but errored to upper layers, causing filesystems to remount read-only, or worse. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_worker.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index 44bf6d1..d48a1df 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -235,7 +235,7 @@ void drbd_endio_pri(struct bio *bio, int error) if (unlikely(error)) { what = (bio_data_dir(bio) == WRITE) ? write_completed_with_error - : (bio_rw(bio) == READA) + : (bio_rw(bio) == READ) ? read_completed_with_error : read_ahead_completed_with_error; } else -- cgit v1.1 From 2395e463fefd4aa8b784787e926e9b84e216d14f Mon Sep 17 00:00:00 2001 From: Randy Dunlap Date: Tue, 11 May 2010 09:02:55 +0200 Subject: paride: fix menu indentation Make the PARIDE menu be displayed correctly, with proper/expected indentation, by moving the GDROM kconfig symbol, which was splitting the PARIDE kconfig symbol from its dependent symbols. Signed-off-by: Randy Dunlap Signed-off-by: Jens Axboe --- drivers/block/Kconfig | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/Kconfig b/drivers/block/Kconfig index 77bfce5..de27768 100644 --- a/drivers/block/Kconfig +++ b/drivers/block/Kconfig @@ -76,6 +76,17 @@ config BLK_DEV_XD It's pretty unlikely that you have one of these: say N. +config GDROM + tristate "SEGA Dreamcast GD-ROM drive" + depends on SH_DREAMCAST + help + A standard SEGA Dreamcast comes with a modified CD ROM drive called a + "GD-ROM" by SEGA to signify it is capable of reading special disks + with up to 1 GB of data. This drive will also read standard CD ROM + disks. Select this option to access any disks in your GD ROM drive. + Most users will want to say "Y" here. + You can also build this as a module which will be called gdrom. + config PARIDE tristate "Parallel port IDE device support" depends on PARPORT_PC @@ -103,17 +114,6 @@ config PARIDE "MicroSolutions backpack protocol", "DataStor Commuter protocol" etc.). -config GDROM - tristate "SEGA Dreamcast GD-ROM drive" - depends on SH_DREAMCAST - help - A standard SEGA Dreamcast comes with a modified CD ROM drive called a - "GD-ROM" by SEGA to signify it is capable of reading special disks - with up to 1 GB of data. This drive will also read standard CD ROM - disks. Select this option to access any disks in your GD ROM drive. - Most users will want to say "Y" here. - You can also build this as a module which will be called gdrom. - source "drivers/block/paride/Kconfig" config BLK_CPQ_DA -- cgit v1.1 From c2d45b4da042720ae94648b417d0e051742441d4 Mon Sep 17 00:00:00 2001 From: Bill Pemberton Date: Fri, 30 Apr 2010 09:34:32 -0400 Subject: cciss: fix shadows sparse warning Fix sparse warnings: drivers/block/cciss.c:1591:37: warning: symbol 'i' shadows an earlier one drivers/block/cciss.c:2437:21: warning: symbol 'i' shadows an earlier one Signed-off-by: Bill Pemberton Acked-by: Mike Miller Signed-off-by: Jiri Kosina --- drivers/block/cciss.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c index eb5ff05..51ceaee 100644 --- a/drivers/block/cciss.c +++ b/drivers/block/cciss.c @@ -1588,7 +1588,6 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, c->Request = ioc->Request; if (ioc->buf_size > 0) { - int i; for (i = 0; i < sg_used; i++) { temp64.val = pci_map_single(host->pdev, buff[i], @@ -2434,7 +2433,7 @@ static int deregister_disk(ctlr_info_t *h, int drv_index, /* if it was the last disk, find the new hightest lun */ if (clear_all && recalculate_highest_lun) { - int i, newhighest = -1; + int newhighest = -1; for (i = 0; i <= h->highest_lun; i++) { /* if the disk has size > 0, it is available */ if (h->drv[i] && h->drv[i]->heads) -- cgit v1.1 From 92183b346f02773dae09182c65f16b013f295d80 Mon Sep 17 00:00:00 2001 From: Geert Uytterhoeven Date: Sun, 5 Apr 2009 13:02:13 +0200 Subject: m68k: amiga - Floppy platform device conversion Signed-off-by: Geert Uytterhoeven --- drivers/block/amiflop.c | 47 +++++++++++++++++++++-------------------------- 1 file changed, 21 insertions(+), 26 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/amiflop.c b/drivers/block/amiflop.c index 0182a22..832798a 100644 --- a/drivers/block/amiflop.c +++ b/drivers/block/amiflop.c @@ -66,6 +66,7 @@ #include #include #include +#include #include #include @@ -1696,34 +1697,18 @@ static struct kobject *floppy_find(dev_t dev, int *part, void *data) return get_disk(unit[drive].gendisk); } -static int __init amiga_floppy_init(void) +static int __init amiga_floppy_probe(struct platform_device *pdev) { int i, ret; - if (!MACH_IS_AMIGA) - return -ENODEV; - - if (!AMIGAHW_PRESENT(AMI_FLOPPY)) - return -ENODEV; - if (register_blkdev(FLOPPY_MAJOR,"fd")) return -EBUSY; - /* - * We request DSKPTR, DSKLEN and DSKDATA only, because the other - * floppy registers are too spreaded over the custom register space - */ - ret = -EBUSY; - if (!request_mem_region(CUSTOM_PHYSADDR+0x20, 8, "amiflop [Paula]")) { - printk("fd: cannot get floppy registers\n"); - goto out_blkdev; - } - ret = -ENOMEM; if ((raw_buf = (char *)amiga_chip_alloc (RAW_BUF_SIZE, "Floppy")) == NULL) { printk("fd: cannot get chip mem buffer\n"); - goto out_memregion; + goto out_blkdev; } ret = -EBUSY; @@ -1792,18 +1777,13 @@ out_irq2: free_irq(IRQ_AMIGA_DSKBLK, NULL); out_irq: amiga_chip_free(raw_buf); -out_memregion: - release_mem_region(CUSTOM_PHYSADDR+0x20, 8); out_blkdev: unregister_blkdev(FLOPPY_MAJOR,"fd"); return ret; } -module_init(amiga_floppy_init); -#ifdef MODULE - #if 0 /* not safe to unload */ -void cleanup_module(void) +static int __exit amiga_floppy_remove(struct platform_device *pdev) { int i; @@ -1820,12 +1800,25 @@ void cleanup_module(void) custom.dmacon = DMAF_DISK; /* disable DMA */ amiga_chip_free(raw_buf); blk_cleanup_queue(floppy_queue); - release_mem_region(CUSTOM_PHYSADDR+0x20, 8); unregister_blkdev(FLOPPY_MAJOR, "fd"); } #endif -#else +static struct platform_driver amiga_floppy_driver = { + .driver = { + .name = "amiga-floppy", + .owner = THIS_MODULE, + }, +}; + +static int __init amiga_floppy_init(void) +{ + return platform_driver_probe(&amiga_floppy_driver, amiga_floppy_probe); +} + +module_init(amiga_floppy_init); + +#ifndef MODULE static int __init amiga_floppy_setup (char *str) { int n; @@ -1840,3 +1833,5 @@ static int __init amiga_floppy_setup (char *str) __setup("floppy=", amiga_floppy_setup); #endif + +MODULE_ALIAS("platform:amiga-floppy"); -- cgit v1.1 From 8c484ee4910b36c9ac273ad1150261c6ebfc1ef7 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Thu, 11 Mar 2010 16:47:58 +0100 Subject: drbd: use proc_create_data with explicit NULL argument To document that we know about deprecation of proc_create, even though we are not affected, as we don't use the ->data member, open code proc_create_data(..., NULL); Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 93d1f9b..db7a07a 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -3129,7 +3129,7 @@ int __init drbd_init(void) if (err) goto Enomem; - drbd_proc = proc_create("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops); + drbd_proc = proc_create_data("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops, NULL); if (!drbd_proc) { printk(KERN_ERR "drbd: unable to register proc file\n"); goto Enomem; -- cgit v1.1 From e4f925e12ea5daaa9baf2dd5af9c4951721dae95 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 17 Mar 2010 14:18:41 +0100 Subject: drbd: Do not upgrade state to Outdated if already Inconsistent [Bugz 277] There was a race condition: In a situation with a SyncSource+Primary and a SyncTarget+Secondary node, and a resync dependency to some other device. After both nodes decided to do the resync, the other device finishes its resync process. At that time SyncSource already sent the P_SYNC_UUID packet, and already updated its peer disk state to Inconsistent. The SyncTarget node waits for the P_SYNC_UUID and sends a state packet to report the resync dependency change. That packet still carries a disk state of Outdated. Impact: If application writes come in, during that time on the Primary node, those do not get replicated, and the out-of-sync counter gets increased. => The completion of resync is not detected on the primary node. => stalled. Those blocks get resync'ed with the next resync, since the are get marked as out-of-sync in the bitmap. In order to fix this, we filter out that wrong state change in the sanitize_state() function. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index db7a07a..8f84a9f 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -840,7 +840,7 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state break; case C_WF_BITMAP_S: case C_PAUSED_SYNC_S: - ns.pdsk = D_OUTDATED; + ns.pdsk = os.pdsk > D_OUTDATED ? D_OUTDATED : os.pdsk; break; case C_SYNC_SOURCE: ns.pdsk = D_INCONSISTENT; -- cgit v1.1 From e0f83012dc510b0be92ee2d59227a573a36777b8 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Thu, 1 Apr 2010 15:13:19 +0200 Subject: drbd: fix regression: attach while connected failed commit e4f925e12ea5daaa9baf2dd5af9c4951721dae95 Author: Philipp Reisner Date: Wed Mar 17 14:18:41 2010 +0100 drbd: Do not upgrade state to Outdated if already Inconsistent prevented the necessary state transition for attaching while connected (Diskless -> Consistent respectively Outdated). This is the fix for the fix. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_main.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 8f84a9f..b1ce5dc 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -840,7 +840,12 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state break; case C_WF_BITMAP_S: case C_PAUSED_SYNC_S: - ns.pdsk = os.pdsk > D_OUTDATED ? D_OUTDATED : os.pdsk; + /* remap any consistent state to D_OUTDATED, + * but disallow "upgrade" of not even consistent states. + */ + ns.pdsk = + (D_DISKLESS < os.pdsk && os.pdsk < D_OUTDATED) + ? os.pdsk : D_OUTDATED; break; case C_SYNC_SOURCE: ns.pdsk = D_INCONSISTENT; -- cgit v1.1 From 8d1894ebe441093cfd967affcbc56b764960575e Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Thu, 1 Apr 2010 16:55:18 +0200 Subject: drbd: remove bogus ASSERT block_id may be ID_SYNCER, as well as checksum based resync request magic, or online verify magic. Let's just drop that ASSERT. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_receiver.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index c786023..93106fb 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -4233,7 +4233,6 @@ static int got_NegRSDReply(struct drbd_conf *mdev, struct p_header *h) sector = be64_to_cpu(p->sector); size = be32_to_cpu(p->blksize); - D_ASSERT(p->block_id == ID_SYNCER); update_peer_seq(mdev, be32_to_cpu(p->seq_num)); -- cgit v1.1 From c3470cde57ea34d9b4bd34891ec040e46b9fb3bf Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Thu, 1 Apr 2010 16:57:19 +0200 Subject: drbd: fix potential protocol error Don't forget to drain the digest in case we cannot satisfy a checksum based resync or online-verify request. It would additionally cause a protocoll error, dropping the connection. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_receiver.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 93106fb..c7285e1 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -1331,6 +1331,9 @@ static int drbd_drain_block(struct drbd_conf *mdev, int data_size) int rr, rv = 1; void *data; + if (!data_size) + return TRUE; + page = drbd_pp_alloc(mdev, 1); data = kmap(page); @@ -1946,7 +1949,7 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) "no local data.\n"); drbd_send_ack_rp(mdev, h->command == P_DATA_REQUEST ? P_NEG_DREPLY : P_NEG_RS_DREPLY , p); - return TRUE; + return drbd_drain_block(mdev, h->length - brps); } /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD -- cgit v1.1 From 8d4ce82b3ccd755c8ba401469ced5286b1e02284 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Thu, 1 Apr 2010 16:59:32 +0200 Subject: drbd: don't start a resync without access to up-to-date Data In case both nodes are "inconsistent", invalidate would have started a resync anyways, without a chance to ever succeed, just filling the logs with warning messages. Simply disallow that state change, re-using the SS_NO_UP_TO_DATE_DISK return value. This also changes the corresponding error string to "Need access to UpToDate Data" -- I found the "Refusing to be Primary without at least one UpToDate disk" answer misleading in some situations anyways. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_main.c | 3 +++ drivers/block/drbd/drbd_strings.c | 2 +- 2 files changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index b1ce5dc..e181e40 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -684,6 +684,9 @@ static int is_valid_state(struct drbd_conf *mdev, union drbd_state ns) else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT) rv = SS_NO_REMOTE_DISK; + else if (ns.conn > C_CONNECTED && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) + rv = SS_NO_UP_TO_DATE_DISK; + else if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S || ns.conn == C_SYNC_SOURCE || diff --git a/drivers/block/drbd/drbd_strings.c b/drivers/block/drbd/drbd_strings.c index 76863e3..85179e1 100644 --- a/drivers/block/drbd/drbd_strings.c +++ b/drivers/block/drbd/drbd_strings.c @@ -70,7 +70,7 @@ static const char *drbd_disk_s_names[] = { static const char *drbd_state_sw_errors[] = { [-SS_TWO_PRIMARIES] = "Multiple primaries not allowed by config", - [-SS_NO_UP_TO_DATE_DISK] = "Refusing to be Primary without at least one UpToDate disk", + [-SS_NO_UP_TO_DATE_DISK] = "Need access to UpToDate data", [-SS_NO_LOCAL_DISK] = "Can not resync without local disk", [-SS_NO_REMOTE_DISK] = "Can not resync without remote disk", [-SS_CONNECTED_OUTDATES] = "Refusing to be Outdated while Connected", -- cgit v1.1 From c3fe30b0e7cd67e0207097f5f39ce9626644879e Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 1 Apr 2010 09:57:40 +0200 Subject: drbd: cleanup: This code path to trigger a resync is no longer needed Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_receiver.c | 17 ----------------- 1 file changed, 17 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index c7285e1..c3504dd 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -2853,7 +2853,6 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) unsigned int max_seg_s; sector_t p_size, p_usize, my_usize; int ldsc = 0; /* local disk size changed */ - enum drbd_conns nconn; ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; if (drbd_recv(mdev, h->payload, h->length) != h->length) @@ -2920,22 +2919,6 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) drbd_set_my_capacity(mdev, p_size); } - if (mdev->p_uuid && mdev->state.conn <= C_CONNECTED && get_ldev(mdev)) { - nconn = drbd_sync_handshake(mdev, - mdev->state.peer, mdev->state.pdsk); - put_ldev(mdev); - - if (nconn == C_MASK) { - drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; - } - - if (drbd_request_state(mdev, NS(conn, nconn)) < SS_SUCCESS) { - drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; - } - } - if (get_ldev(mdev)) { if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) { mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); -- cgit v1.1 From 6666032ade5a758aa05380ab92f416ab8ef25005 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Tue, 6 Apr 2010 12:15:04 +0200 Subject: drbd: check for corrupt or malicous sector addresses when receiving data Even if it should never happen if the peer does behave, we need to double check, and not even attempt access beyond end of device. It usually would be caught by lower layers, resulting in "IO error", but may also end up in the internal meta data area. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_receiver.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index c3504dd..3a36bc8 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -1262,6 +1262,7 @@ static int receive_Barrier(struct drbd_conf *mdev, struct p_header *h) static struct drbd_epoch_entry * read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __must_hold(local) { + const sector_t capacity = drbd_get_capacity(mdev->this_bdev); struct drbd_epoch_entry *e; struct bio_vec *bvec; struct page *page; @@ -1287,6 +1288,15 @@ read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __ ERR_IF(data_size & 0x1ff) return NULL; ERR_IF(data_size > DRBD_MAX_SEGMENT_SIZE) return NULL; + /* even though we trust out peer, + * we sometimes have to double check. */ + if (sector + (data_size>>9) > capacity) { + dev_err(DEV, "capacity: %llus < sector: %llus + size: %u\n", + (unsigned long long)capacity, + (unsigned long long)sector, data_size); + return NULL; + } + /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD * "criss-cross" setup, that might cause write-out on some other DRBD, * which in turn might block on the other node at this very place. */ -- cgit v1.1 From 979f5c7f1f6c8a532b943defb790d43b999934eb Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Tue, 6 Apr 2010 14:15:06 +0200 Subject: drbd: fail_requests_early: remove incorrect and unnecessary optimization The condition does not fit the commend (I may well be Primary, even if I lost the disk earlier and now the connection). And this is catched below anyways, where it also gets logged. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_req.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c index de81ab7..d8d9bbf 100644 --- a/drivers/block/drbd/drbd_req.c +++ b/drivers/block/drbd/drbd_req.c @@ -962,11 +962,6 @@ fail_and_free_req: */ static int drbd_fail_request_early(struct drbd_conf *mdev, int is_write) { - /* Unconfigured */ - if (mdev->state.conn == C_DISCONNECTING && - mdev->state.disk == D_DISKLESS) - return 1; - if (mdev->state.role != R_PRIMARY && (!allow_oos || is_write)) { if (__ratelimit(&drbd_ratelimit_state)) { -- cgit v1.1 From 3a11a4878939e0e3c355bf3f52ef642a4cb6ba84 Mon Sep 17 00:00:00 2001 From: Adam Gandelman Date: Thu, 8 Apr 2010 16:48:23 -0700 Subject: drbd: New handler: initial-split-brain Some wish to be notified of all instances of split brain, not just those that go unresolved. The initial-split-brain handler is called to notify someone upon detection of all split brain conditions even if auto-recovery policies are configured. Signed-off-by: Adam Gandelman Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_receiver.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 3a36bc8..6876041 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -2487,6 +2487,9 @@ static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_rol hg > 0 ? "source" : "target"); } + if (abs(hg) == 100) + drbd_khelper(mdev, "initial-split-brain"); + if (hg == 100 || (hg == -100 && mdev->net_conf->always_asbp)) { int pcount = (mdev->state.role == R_PRIMARY) + (peer_role == R_PRIMARY); @@ -2532,7 +2535,7 @@ static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_rol * after an attempted attach on a diskless node. * We just refuse to attach -- well, we drop the "connection" * to that disk, in a way... */ - dev_alert(DEV, "Split-Brain detected, dropping connection!\n"); + dev_alert(DEV, "Split-Brain detected but unresolved, dropping connection!\n"); drbd_khelper(mdev, "split-brain"); return C_MASK; } -- cgit v1.1 From d845030f21859dd11bcecc7e1b8575fb845eb425 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 24 Mar 2010 15:51:26 +0100 Subject: drbd: made determin_dev_size's parameter an flag enum Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 6 +++++- drivers/block/drbd/drbd_nl.c | 6 +++--- 2 files changed, 8 insertions(+), 4 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index d6f1ae3..f5c56f4 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -1382,8 +1382,12 @@ extern void drbd_suspend_io(struct drbd_conf *mdev); extern void drbd_resume_io(struct drbd_conf *mdev); extern char *ppsize(char *buf, unsigned long long size); extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int); +enum dds_flags { + DDSF_FORCED = 1, + DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ +}; enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 }; -extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, int force) __must_hold(local); +extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local); extern void resync_after_online_grow(struct drbd_conf *); extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local); extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 6429d2b..97abbc2 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -510,7 +510,7 @@ void drbd_resume_io(struct drbd_conf *mdev) * Returns 0 on success, negative return values indicate errors. * You should call drbd_md_sync() after calling this function. */ -enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, int force) __must_hold(local) +enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local) { sector_t prev_first_sect, prev_size; /* previous meta location */ sector_t la_size; @@ -541,7 +541,7 @@ enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, int force /* TODO: should only be some assert here, not (re)init... */ drbd_md_set_sector_offsets(mdev, mdev->ldev); - size = drbd_new_dev_size(mdev, mdev->ldev, force); + size = drbd_new_dev_size(mdev, mdev->ldev, flags & DDSF_FORCED); if (drbd_get_capacity(mdev->this_bdev) != size || drbd_bm_capacity(mdev) != size) { @@ -1508,7 +1508,7 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, } mdev->ldev->dc.disk_size = (sector_t)rs.resize_size; - dd = drbd_determin_dev_size(mdev, rs.resize_force); + dd = drbd_determin_dev_size(mdev, rs.resize_force ? DDSF_FORCED : 0); drbd_md_sync(mdev); put_ldev(mdev); if (dd == dev_size_error) { -- cgit v1.1 From 02d9a94bbb0d4e0fec8db6735bdc4ccfaac8f0ce Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 24 Mar 2010 16:23:03 +0100 Subject: drbd: Implemented the set_new_bits parameter for drbd_bm_resize() Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 10 +++++++--- drivers/block/drbd/drbd_int.h | 2 +- drivers/block/drbd/drbd_main.c | 2 +- drivers/block/drbd/drbd_nl.c | 2 +- 4 files changed, 10 insertions(+), 6 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 3390716..695fb64 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -441,7 +441,7 @@ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) * In case this is actually a resize, we copy the old bitmap into the new one. * Otherwise, the bitmap is initialized to all bits set. */ -int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity) +int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) { struct drbd_bitmap *b = mdev->bitmap; unsigned long bits, words, owords, obits, *p_addr, *bm; @@ -526,8 +526,12 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity) b->bm_dev_capacity = capacity; if (growing) { - bm_memset(b, owords, 0xff, words-owords); - b->bm_set += bits - obits; + if (set_new_bits) { + bm_memset(b, owords, 0xff, words-owords); + b->bm_set += bits - obits; + } else + bm_memset(b, owords, 0x00, words-owords); + } if (want < have) { diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index f5c56f4..37a25a6 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -1311,7 +1311,7 @@ struct bm_extent { #define APP_R_HSIZE 15 extern int drbd_bm_init(struct drbd_conf *mdev); -extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors); +extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits); extern void drbd_bm_cleanup(struct drbd_conf *mdev); extern void drbd_bm_set_all(struct drbd_conf *mdev); extern void drbd_bm_clear_all(struct drbd_conf *mdev); diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index e181e40..65c2a65 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2688,7 +2688,7 @@ void drbd_mdev_cleanup(struct drbd_conf *mdev) drbd_set_my_capacity(mdev, 0); if (mdev->bitmap) { /* maybe never allocated. */ - drbd_bm_resize(mdev, 0); + drbd_bm_resize(mdev, 0, 1); drbd_bm_cleanup(mdev); } diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 97abbc2..360e506 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -546,7 +546,7 @@ enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, enum dds_ if (drbd_get_capacity(mdev->this_bdev) != size || drbd_bm_capacity(mdev) != size) { int err; - err = drbd_bm_resize(mdev, size); + err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC)); if (unlikely(err)) { /* currently there is only one error: ENOMEM! */ size = drbd_bm_capacity(mdev)>>1; -- cgit v1.1 From e89b591c3aba0af87f5248b15f56ce7a4f439c16 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 24 Mar 2010 17:11:33 +0100 Subject: drbd: Implemented flags for the resize packet Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 14 ++++++++------ drivers/block/drbd/drbd_main.c | 8 ++++---- drivers/block/drbd/drbd_nl.c | 2 +- drivers/block/drbd/drbd_receiver.c | 19 ++++++++++++------- 4 files changed, 25 insertions(+), 18 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 37a25a6..e091324 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -481,7 +481,8 @@ struct p_sizes { u64 u_size; /* user requested size */ u64 c_size; /* current exported size */ u32 max_segment_size; /* Maximal size of a BIO */ - u32 queue_order_type; + u16 queue_order_type; /* not yet implemented in DRBD*/ + u16 dds_flags; /* use enum dds_flags here. */ } __packed; struct p_state { @@ -1081,6 +1082,11 @@ enum chg_state_flags { CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE, }; +enum dds_flags { + DDSF_FORCED = 1, + DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ +}; + extern void drbd_init_set_defaults(struct drbd_conf *mdev); extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f, union drbd_state mask, union drbd_state val); @@ -1113,7 +1119,7 @@ extern int drbd_send_protocol(struct drbd_conf *mdev); extern int drbd_send_uuids(struct drbd_conf *mdev); extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev); extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val); -extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply); +extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags); extern int _drbd_send_state(struct drbd_conf *mdev); extern int drbd_send_state(struct drbd_conf *mdev); extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, @@ -1382,10 +1388,6 @@ extern void drbd_suspend_io(struct drbd_conf *mdev); extern void drbd_resume_io(struct drbd_conf *mdev); extern char *ppsize(char *buf, unsigned long long size); extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int); -enum dds_flags { - DDSF_FORCED = 1, - DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ -}; enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 }; extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local); extern void resync_after_online_grow(struct drbd_conf *); diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 65c2a65..a478dad 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1240,7 +1240,7 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) { kfree(mdev->p_uuid); /* We expect to receive up-to-date UUIDs soon. */ mdev->p_uuid = NULL; /* ...to not use the old ones in the mean time */ - drbd_send_sizes(mdev, 0); /* to start sync... */ + drbd_send_sizes(mdev, 0, 0); /* to start sync... */ drbd_send_uuids(mdev); drbd_send_state(mdev); } @@ -1763,7 +1763,7 @@ int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val) (struct p_header *)&p, sizeof(p)); } -int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply) +int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags) { struct p_sizes p; sector_t d_size, u_size; @@ -1775,7 +1775,6 @@ int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply) d_size = drbd_get_max_capacity(mdev->ldev); u_size = mdev->ldev->dc.disk_size; q_order_type = drbd_queue_order_type(mdev); - p.queue_order_type = cpu_to_be32(drbd_queue_order_type(mdev)); put_ldev(mdev); } else { d_size = 0; @@ -1787,7 +1786,8 @@ int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply) p.u_size = cpu_to_be64(u_size); p.c_size = cpu_to_be64(trigger_reply ? 0 : drbd_get_capacity(mdev->this_bdev)); p.max_segment_size = cpu_to_be32(queue_max_segment_size(mdev->rq_queue)); - p.queue_order_type = cpu_to_be32(q_order_type); + p.queue_order_type = cpu_to_be16(q_order_type); + p.dds_flags = cpu_to_be16(flags); ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SIZES, (struct p_header *)&p, sizeof(p)); diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 360e506..6f79333 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -1521,7 +1521,7 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, set_bit(RESIZE_PENDING, &mdev->flags); drbd_send_uuids(mdev); - drbd_send_sizes(mdev, 1); + drbd_send_sizes(mdev, 1, 0); } fail: diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 6876041..11b1baf 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -902,7 +902,7 @@ retry: if (!drbd_send_protocol(mdev)) return -1; drbd_send_sync_param(mdev, &mdev->sync_conf); - drbd_send_sizes(mdev, 0); + drbd_send_sizes(mdev, 0, 0); drbd_send_uuids(mdev); drbd_send_state(mdev); clear_bit(USE_DEGR_WFC_T, &mdev->flags); @@ -2866,6 +2866,7 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) unsigned int max_seg_s; sector_t p_size, p_usize, my_usize; int ldsc = 0; /* local disk size changed */ + enum dds_flags ddsf; ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; if (drbd_recv(mdev, h->payload, h->length) != h->length) @@ -2921,8 +2922,9 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) } #undef min_not_zero + ddsf = be16_to_cpu(p->dds_flags); if (get_ldev(mdev)) { - dd = drbd_determin_dev_size(mdev, 0); + dd = drbd_determin_dev_size(mdev, ddsf); put_ldev(mdev); if (dd == dev_size_error) return FALSE; @@ -2942,7 +2944,7 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) if (max_seg_s != queue_max_segment_size(mdev->rq_queue)) drbd_setup_queue_param(mdev, max_seg_s); - drbd_setup_order_type(mdev, be32_to_cpu(p->queue_order_type)); + drbd_setup_order_type(mdev, be16_to_cpu(p->queue_order_type)); put_ldev(mdev); } @@ -2951,14 +2953,17 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) drbd_get_capacity(mdev->this_bdev) || ldsc) { /* we have different sizes, probably peer * needs to know my new size... */ - drbd_send_sizes(mdev, 0); + drbd_send_sizes(mdev, 0, ddsf); } if (test_and_clear_bit(RESIZE_PENDING, &mdev->flags) || (dd == grew && mdev->state.conn == C_CONNECTED)) { if (mdev->state.pdsk >= D_INCONSISTENT && - mdev->state.disk >= D_INCONSISTENT) - resync_after_online_grow(mdev); - else + mdev->state.disk >= D_INCONSISTENT) { + if (ddsf & DDSF_NO_RESYNC) + dev_info(DEV, "Resync of new storage suppressed with --assume-clean\n"); + else + resync_after_online_grow(mdev); + } else set_bit(RESYNC_AFTER_NEG, &mdev->flags); } } -- cgit v1.1 From fd76438c2421324fa2fb9303e760ec5332ff0b58 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 1 Apr 2010 09:57:40 +0200 Subject: drbd: Make sure to resync all of the new storage upon online resize Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 695fb64..178cf16 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -84,6 +84,9 @@ struct drbd_bitmap { #define BM_MD_IO_ERROR 1 #define BM_P_VMALLOCED 2 +int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, + unsigned long e, int val, const enum km_type km); + static int bm_is_locked(struct drbd_bitmap *b) { return test_bit(BM_LOCKED, &b->bm_flags); @@ -529,6 +532,9 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) if (set_new_bits) { bm_memset(b, owords, 0xff, words-owords); b->bm_set += bits - obits; + __bm_change_bits_to(mdev, obits, + ALIGN(obits, BITS_PER_LONG), + 1, KM_IRQ1); } else bm_memset(b, owords, 0x00, words-owords); -- cgit v1.1 From b4ee79dac3bddc468e21cae0deb00b80ec4ac051 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 1 Apr 2010 09:57:40 +0200 Subject: drbd: Added some missing statics Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 178cf16..aa7e23c 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -84,7 +84,7 @@ struct drbd_bitmap { #define BM_MD_IO_ERROR 1 #define BM_P_VMALLOCED 2 -int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, +static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, unsigned long e, int val, const enum km_type km); static int bm_is_locked(struct drbd_bitmap *b) @@ -783,7 +783,7 @@ static void bm_page_io_async(struct drbd_conf *mdev, struct drbd_bitmap *b, int /* nothing to do, on disk == in memory */ # define bm_cpu_to_lel(x) ((void)0) # else -void bm_cpu_to_lel(struct drbd_bitmap *b) +static void bm_cpu_to_lel(struct drbd_bitmap *b) { /* need to cpu_to_lel all the pages ... * this may be optimized by using @@ -1025,7 +1025,7 @@ unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_f * wants bitnr, not sector. * expected to be called for only a few bits (e - s about BITS_PER_LONG). * Must hold bitmap lock already. */ -int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, +static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, unsigned long e, int val, const enum km_type km) { struct drbd_bitmap *b = mdev->bitmap; @@ -1063,7 +1063,7 @@ int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, * for val != 0, we change 0 -> 1, return code positive * for val == 0, we change 1 -> 0, return code negative * wants bitnr, not sector */ -int bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, +static int bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, const unsigned long e, int val) { unsigned long flags; -- cgit v1.1 From 6495d2c6d04f4c45411fdb1b40527c24015f39d6 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 24 Mar 2010 16:07:04 +0100 Subject: drbd: Implemented the --assume-clean option for drbdsetup resize Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_nl.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 6f79333..19b9a28 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -1479,6 +1479,7 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, int retcode = NO_ERROR; int ldsc = 0; /* local disk size changed */ enum determine_dev_size dd; + enum dds_flags ddsf; memset(&rs, 0, sizeof(struct resize)); if (!resize_from_tags(mdev, nlp->tag_list, &rs)) { @@ -1502,13 +1503,19 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, goto fail; } + if (rs.no_resync && mdev->agreed_pro_version < 93) { + retcode = ERR_NEED_APV_93; + goto fail; + } + if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) { mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); ldsc = 1; } mdev->ldev->dc.disk_size = (sector_t)rs.resize_size; - dd = drbd_determin_dev_size(mdev, rs.resize_force ? DDSF_FORCED : 0); + ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); + dd = drbd_determin_dev_size(mdev, ddsf); drbd_md_sync(mdev); put_ldev(mdev); if (dd == dev_size_error) { @@ -1521,7 +1528,7 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, set_bit(RESIZE_PENDING, &mdev->flags); drbd_send_uuids(mdev); - drbd_send_sizes(mdev, 1, 0); + drbd_send_sizes(mdev, 1, ddsf); } fail: -- cgit v1.1 From 087c24925cf4209be1a91f8ede9241e17e9734c7 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Fri, 26 Mar 2010 13:49:56 +0100 Subject: drbd: bugfix: Make resize work, if remote's size was limiting and increased in the meantime Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_nl.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 19b9a28..6cb7036 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -1477,7 +1477,6 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, { struct resize rs; int retcode = NO_ERROR; - int ldsc = 0; /* local disk size changed */ enum determine_dev_size dd; enum dds_flags ddsf; @@ -1508,10 +1507,8 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, goto fail; } - if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) { + if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); - ldsc = 1; - } mdev->ldev->dc.disk_size = (sector_t)rs.resize_size; ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); @@ -1523,7 +1520,7 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, goto fail; } - if (mdev->state.conn == C_CONNECTED && (dd != unchanged || ldsc)) { + if (mdev->state.conn == C_CONNECTED) { if (dd == grew) set_bit(RESIZE_PENDING, &mdev->flags); -- cgit v1.1 From 6b4388ac1f282515db3a651707238cad00b50e80 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Mon, 26 Apr 2010 14:11:45 +0200 Subject: drbd: Added transmission faults to the fault injection code Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 1 + drivers/block/drbd/drbd_main.c | 3 ++- drivers/block/drbd/drbd_receiver.c | 8 +++++++- 3 files changed, 10 insertions(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index e091324..2409de1 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -132,6 +132,7 @@ enum { DRBD_FAULT_DT_RA = 6, /* data read ahead */ DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ DRBD_FAULT_AL_EE = 8, /* alloc ee */ + DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ DRBD_FAULT_MAX, }; diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index a478dad..7468d2c 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -3668,7 +3668,8 @@ _drbd_fault_str(unsigned int type) { [DRBD_FAULT_DT_RD] = "Data read", [DRBD_FAULT_DT_RA] = "Data read ahead", [DRBD_FAULT_BM_ALLOC] = "BM allocation", - [DRBD_FAULT_AL_EE] = "EE allocation" + [DRBD_FAULT_AL_EE] = "EE allocation", + [DRBD_FAULT_RECEIVE] = "receive data corruption", }; return (type < DRBD_FAULT_MAX) ? _faults[type] : "**Unknown**"; diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 11b1baf..b27f4dd 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -1270,6 +1270,7 @@ read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __ int dgs, ds, i, rr; void *dig_in = mdev->int_dig_in; void *dig_vv = mdev->int_dig_vv; + unsigned long *data; dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_r_tfm) ? crypto_hash_digestsize(mdev->integrity_r_tfm) : 0; @@ -1307,7 +1308,12 @@ read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __ ds = data_size; bio_for_each_segment(bvec, bio, i) { page = bvec->bv_page; - rr = drbd_recv(mdev, kmap(page), min_t(int, ds, PAGE_SIZE)); + data = kmap(page); + rr = drbd_recv(mdev, data, min_t(int, ds, PAGE_SIZE)); + if (FAULT_ACTIVE(mdev, DRBD_FAULT_RECEIVE)) { + dev_err(DEV, "Fault injection: Corrupting data on receive\n"); + data[0] = data[0] ^ (unsigned long)-1; + } kunmap(page); if (rr != min_t(int, ds, PAGE_SIZE)) { drbd_free_ee(mdev, e); -- cgit v1.1 From 5223671bb0315d83f9ad7becbbb9e703aa735bbe Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 28 Apr 2010 14:46:57 +0200 Subject: drbd: Fixed bitmap in case of online-grow without resync The "surplus" bits of the old (smaller) bitmap must be clean in case of online-grow without resync. Note: Reverted 67ae8b80d4a116ab3b7094eb3723506b20c06dff as well, since the lines added by this patch are redundant. The bits get set by the bm_set_surplus(b) call before that. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index aa7e23c..e3f88d6 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -519,7 +519,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) obits = b->bm_bits; growing = bits > obits; - if (opages) + if (opages && growing && set_new_bits) bm_set_surplus(b); b->bm_pages = npages; @@ -532,9 +532,6 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) if (set_new_bits) { bm_memset(b, owords, 0xff, words-owords); b->bm_set += bits - obits; - __bm_change_bits_to(mdev, obits, - ALIGN(obits, BITS_PER_LONG), - 1, KM_IRQ1); } else bm_memset(b, owords, 0x00, words-owords); -- cgit v1.1 From 0ced55a3bed25b0e30dcb3c7dce9634ce3c60cf2 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Fri, 30 Apr 2010 15:26:20 +0200 Subject: drbd: Receiving of delay_probes Delay_probes are new packets in the DRBD protocol, which allow DRBD to know the current delay packets have on the data socket. (relative to the meta data socket) Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 20 +++++++- drivers/block/drbd/drbd_main.c | 3 ++ drivers/block/drbd/drbd_receiver.c | 96 ++++++++++++++++++++++++++++++++++++++ 3 files changed, 118 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 2409de1..fd7615f 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -209,8 +209,11 @@ enum drbd_packets { P_RS_IS_IN_SYNC = 0x22, /* meta socket */ P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */ P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */ + /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */ + /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */ + P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */ - P_MAX_CMD = 0x25, + P_MAX_CMD = 0x28, P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */ P_MAX_OPT_CMD = 0x101, @@ -540,6 +543,18 @@ struct p_compressed_bm { u8 code[0]; } __packed; +struct p_delay_probe { + struct p_header head; + u32 seq_num; /* sequence number to match the two probe packets */ + u32 offset; /* usecs the probe got sent after the reference time point */ +} __packed; + +struct delay_probe { + struct list_head list; + int seq_num; + struct timeval time; +}; + /* DCBP: Drbd Compressed Bitmap Packet ... */ static inline enum drbd_bitmap_code DCBP_get_code(struct p_compressed_bm *p) @@ -1028,6 +1043,9 @@ struct drbd_conf { u64 ed_uuid; /* UUID of the exposed data */ struct mutex state_mutex; char congestion_reason; /* Why we where congested... */ + struct list_head delay_probes; /* protected by peer_seq_lock */ + int data_delay; /* Delay of packets on the data-sock behind meta-sock */ + atomic_t delay_seq; /* To generate sequence numbers of delay probes */ }; static inline struct drbd_conf *minor_to_mdev(unsigned int minor) diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 7468d2c..3d5fe30 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2608,6 +2608,7 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) atomic_set(&mdev->net_cnt, 0); atomic_set(&mdev->packet_seq, 0); atomic_set(&mdev->pp_in_use, 0); + atomic_set(&mdev->delay_seq, 0); mutex_init(&mdev->md_io_mutex); mutex_init(&mdev->data.mutex); @@ -2636,6 +2637,8 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) INIT_LIST_HEAD(&mdev->unplug_work.list); INIT_LIST_HEAD(&mdev->md_sync_work.list); INIT_LIST_HEAD(&mdev->bm_io_work.w.list); + INIT_LIST_HEAD(&mdev->delay_probes); + mdev->resync_work.cb = w_resync_inactive; mdev->unplug_work.cb = w_send_write_hint; mdev->md_sync_work.cb = w_md_sync; diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index b27f4dd..fee0d24 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -3501,6 +3501,92 @@ static int receive_UnplugRemote(struct drbd_conf *mdev, struct p_header *h) return TRUE; } +static void timeval_sub_us(struct timeval* tv, unsigned int us) +{ + tv->tv_sec -= us / 1000000; + us = us % 1000000; + if (tv->tv_usec > us) { + tv->tv_usec += 1000000; + tv->tv_sec--; + } + tv->tv_usec -= us; +} + +static void got_delay_probe(struct drbd_conf *mdev, int from, struct p_delay_probe *p) +{ + struct delay_probe *dp; + struct list_head *le; + struct timeval now; + int seq_num; + int offset; + int data_delay; + + seq_num = be32_to_cpu(p->seq_num); + offset = be32_to_cpu(p->offset); + + spin_lock(&mdev->peer_seq_lock); + if (!list_empty(&mdev->delay_probes)) { + if (from == USE_DATA_SOCKET) + le = mdev->delay_probes.next; + else + le = mdev->delay_probes.prev; + + dp = list_entry(le, struct delay_probe, list); + + if (dp->seq_num == seq_num) { + list_del(le); + spin_unlock(&mdev->peer_seq_lock); + do_gettimeofday(&now); + timeval_sub_us(&now, offset); + data_delay = + now.tv_usec - dp->time.tv_usec + + (now.tv_sec - dp->time.tv_sec) * 1000000; + + if (data_delay > 0) + mdev->data_delay = data_delay; + + kfree(dp); + return; + } + + if (dp->seq_num > seq_num) { + spin_unlock(&mdev->peer_seq_lock); + dev_warn(DEV, "Previous allocation failure of struct delay_probe?\n"); + return; /* Do not alloca a struct delay_probe.... */ + } + } + spin_unlock(&mdev->peer_seq_lock); + + dp = kmalloc(sizeof(struct delay_probe), GFP_NOIO); + if (!dp) { + dev_warn(DEV, "Failed to allocate a struct delay_probe, do not worry.\n"); + return; + } + + dp->seq_num = seq_num; + do_gettimeofday(&dp->time); + timeval_sub_us(&dp->time, offset); + + spin_lock(&mdev->peer_seq_lock); + if (from == USE_DATA_SOCKET) + list_add(&dp->list, &mdev->delay_probes); + else + list_add_tail(&dp->list, &mdev->delay_probes); + spin_unlock(&mdev->peer_seq_lock); +} + +static int receive_delay_probe(struct drbd_conf *mdev, struct p_header *h) +{ + struct p_delay_probe *p = (struct p_delay_probe *)h; + + ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; + if (drbd_recv(mdev, h->payload, h->length) != h->length) + return FALSE; + + got_delay_probe(mdev, USE_DATA_SOCKET, p); + return TRUE; +} + typedef int (*drbd_cmd_handler_f)(struct drbd_conf *, struct p_header *); static drbd_cmd_handler_f drbd_default_handler[] = { @@ -3524,6 +3610,7 @@ static drbd_cmd_handler_f drbd_default_handler[] = { [P_OV_REQUEST] = receive_DataRequest, [P_OV_REPLY] = receive_DataRequest, [P_CSUM_RS_REQUEST] = receive_DataRequest, + [P_DELAY_PROBE] = receive_delay_probe, /* anything missing from this table is in * the asender_tbl, see get_asender_cmd */ [P_MAX_CMD] = NULL, @@ -4300,6 +4387,14 @@ static int got_OVResult(struct drbd_conf *mdev, struct p_header *h) return TRUE; } +static int got_delay_probe_m(struct drbd_conf *mdev, struct p_header *h) +{ + struct p_delay_probe *p = (struct p_delay_probe *)h; + + got_delay_probe(mdev, USE_META_SOCKET, p); + return TRUE; +} + struct asender_cmd { size_t pkt_size; int (*process)(struct drbd_conf *mdev, struct p_header *h); @@ -4324,6 +4419,7 @@ static struct asender_cmd *get_asender_cmd(int cmd) [P_BARRIER_ACK] = { sizeof(struct p_barrier_ack), got_BarrierAck }, [P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), got_RqSReply }, [P_RS_IS_IN_SYNC] = { sizeof(struct p_block_ack), got_IsInSync }, + [P_DELAY_PROBE] = { sizeof(struct p_delay_probe), got_delay_probe_m }, [P_MAX_CMD] = { 0, NULL }, }; if (cmd > P_MAX_CMD || asender_tbl[cmd].process == NULL) -- cgit v1.1 From 7237bc430f49de1145d761c4b39f2ebae58842d5 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Mon, 3 May 2010 15:10:47 +0200 Subject: drbd: Sending of delay_probes Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 1 + drivers/block/drbd/drbd_main.c | 33 +++++++++++++++++++++++++++++++++ 2 files changed, 34 insertions(+) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index fd7615f..3e4d8b5 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -1046,6 +1046,7 @@ struct drbd_conf { struct list_head delay_probes; /* protected by peer_seq_lock */ int data_delay; /* Delay of packets on the data-sock behind meta-sock */ atomic_t delay_seq; /* To generate sequence numbers of delay probes */ + struct timeval dps_time; /* delay-probes-start-time */ }; static inline struct drbd_conf *minor_to_mdev(unsigned int minor) diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 3d5fe30..710bfeb 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2188,6 +2188,39 @@ int drbd_send_ov_request(struct drbd_conf *mdev, sector_t sector, int size) return ok; } +static int drbd_send_delay_probe(struct drbd_conf *mdev, struct drbd_socket *ds) +{ + struct p_delay_probe dp; + int offset, ok = 0; + struct timeval now; + + mutex_lock(&ds->mutex); + if (likely(ds->socket)) { + do_gettimeofday(&now); + offset = now.tv_usec - mdev->dps_time.tv_usec + + (now.tv_sec - mdev->dps_time.tv_sec) * 1000000; + dp.seq_num = cpu_to_be32(atomic_read(&mdev->delay_seq)); + dp.offset = cpu_to_be32(offset); + + ok = _drbd_send_cmd(mdev, ds->socket, P_DELAY_PROBE, + (struct p_header *)&dp, sizeof(dp), 0); + } + mutex_unlock(&ds->mutex); + + return ok; +} + +static int drbd_send_dalay_probes(struct drbd_conf *mdev) +{ + int ok; + atomic_inc(&mdev->delay_seq); + do_gettimeofday(&mdev->dps_time); + ok = drbd_send_delay_probe(mdev, &mdev->meta); + ok = ok && drbd_send_delay_probe(mdev, &mdev->data); + + return ok; +} + /* called on sndtimeo * returns FALSE if we should retry, * TRUE if we think connection is dead -- cgit v1.1 From 67c7ddd055c794f0d8e9466ca2d6b5cc0b73d4df Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Tue, 4 May 2010 11:12:00 +0200 Subject: drbd: Four new configuration settings for resync speed control To reasonably control resync speed over drbd-proxy connections, drbd has to measure the current delay of packets transmitted over the (possibly congested) data socket vs the meta-data socket. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_nl.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 6cb7036..93d1506 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -1555,6 +1555,10 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n sc.rate = DRBD_RATE_DEF; sc.after = DRBD_AFTER_DEF; sc.al_extents = DRBD_AL_EXTENTS_DEF; + sc.dp_volume = DRBD_DP_VOLUME_DEF; + sc.dp_interval = DRBD_DP_INTERVAL_DEF; + sc.throttle_th = DRBD_RS_THROTTLE_TH_DEF; + sc.hold_off_th = DRBD_RS_HOLD_OFF_TH_DEF; } else memcpy(&sc, &mdev->sync_conf, sizeof(struct syncer_conf)); -- cgit v1.1 From bd26bfc5b4253425d17aa49648ae1f3e976041c4 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Tue, 4 May 2010 12:33:58 +0200 Subject: drbd: Actually send delay probes Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 6 +++++- drivers/block/drbd/drbd_main.c | 43 +++++++++++++++++++++++++++++++++++++++++- 2 files changed, 47 insertions(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 3e4d8b5..210870e 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -925,9 +925,11 @@ struct drbd_conf { unsigned int ko_count; struct drbd_work resync_work, unplug_work, - md_sync_work; + md_sync_work, + delay_probe_work; struct timer_list resync_timer; struct timer_list md_sync_timer; + struct timer_list delay_probe_timer; /* Used after attach while negotiating new disk state. */ union drbd_state new_state_tmp; @@ -1047,6 +1049,8 @@ struct drbd_conf { int data_delay; /* Delay of packets on the data-sock behind meta-sock */ atomic_t delay_seq; /* To generate sequence numbers of delay probes */ struct timeval dps_time; /* delay-probes-start-time */ + int dp_volume_last; /* send_cnt of last delay probe */ + int c_sync_rate; /* current resync rate after delay_probe magic */ }; static inline struct drbd_conf *minor_to_mdev(unsigned int minor) diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 710bfeb..98785d0 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2207,10 +2207,13 @@ static int drbd_send_delay_probe(struct drbd_conf *mdev, struct drbd_socket *ds) } mutex_unlock(&ds->mutex); + mdev->dp_volume_last = mdev->send_cnt; + mod_timer(&mdev->delay_probe_timer, jiffies + mdev->sync_conf.dp_interval * HZ / 10); + return ok; } -static int drbd_send_dalay_probes(struct drbd_conf *mdev) +static int drbd_send_delay_probes(struct drbd_conf *mdev) { int ok; atomic_inc(&mdev->delay_seq); @@ -2350,6 +2353,30 @@ static int _drbd_send_zc_bio(struct drbd_conf *mdev, struct bio *bio) return 1; } +static void consider_delay_probes(struct drbd_conf *mdev) +{ + if (mdev->state.conn != C_SYNC_SOURCE) + return; + + if (mdev->dp_volume_last + mdev->sync_conf.dp_volume * 2 < mdev->send_cnt) + drbd_send_delay_probes(mdev); +} + +static int w_delay_probes(struct drbd_conf *mdev, struct drbd_work *w, int cancel) +{ + if (!cancel && mdev->state.conn == C_SYNC_SOURCE) + drbd_send_delay_probes(mdev); + + return 1; +} + +static void delay_probe_timer_fn(unsigned long data) +{ + struct drbd_conf *mdev = (struct drbd_conf *) data; + + drbd_queue_work(&mdev->data.work, &mdev->delay_probe_work); +} + /* Used to send write requests * R_PRIMARY -> Peer (P_DATA) */ @@ -2412,6 +2439,10 @@ int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req) } drbd_put_data_sock(mdev); + + if (ok) + consider_delay_probes(mdev); + return ok; } @@ -2457,6 +2488,10 @@ int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, ok = _drbd_send_zc_bio(mdev, e->private_bio); drbd_put_data_sock(mdev); + + if (ok) + consider_delay_probes(mdev); + return ok; } @@ -2671,17 +2706,23 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) INIT_LIST_HEAD(&mdev->md_sync_work.list); INIT_LIST_HEAD(&mdev->bm_io_work.w.list); INIT_LIST_HEAD(&mdev->delay_probes); + INIT_LIST_HEAD(&mdev->delay_probe_work.list); mdev->resync_work.cb = w_resync_inactive; mdev->unplug_work.cb = w_send_write_hint; mdev->md_sync_work.cb = w_md_sync; mdev->bm_io_work.w.cb = w_bitmap_io; + mdev->delay_probe_work.cb = w_delay_probes; init_timer(&mdev->resync_timer); init_timer(&mdev->md_sync_timer); + init_timer(&mdev->delay_probe_timer); mdev->resync_timer.function = resync_timer_fn; mdev->resync_timer.data = (unsigned long) mdev; mdev->md_sync_timer.function = md_sync_timer_fn; mdev->md_sync_timer.data = (unsigned long) mdev; + mdev->delay_probe_timer.function = delay_probe_timer_fn; + mdev->delay_probe_timer.data = (unsigned long) mdev; + init_waitqueue_head(&mdev->misc_wait); init_waitqueue_head(&mdev->state_wait); -- cgit v1.1 From cdd67a74603d0453ddffc24c572aed2ddd1795b8 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Tue, 4 May 2010 16:57:18 +0200 Subject: drbd: Control the actual resync rate based on the queuing delay of data packets In a setup with a high bandwidth and high latency network, eventually involving deep queues in routers, it is beneficial to only fill those queues up to an limited extend with resync data. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_worker.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index 44bf6d1..0bbecf4 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -414,6 +414,18 @@ void resync_timer_fn(unsigned long data) drbd_queue_work(&mdev->data.work, &mdev->resync_work); } +static int calc_resync_rate(struct drbd_conf *mdev) +{ + int d = mdev->data_delay / 1000; /* us -> ms */ + int td = mdev->sync_conf.throttle_th * 100; /* 0.1s -> ms */ + int hd = mdev->sync_conf.hold_off_th * 100; /* 0.1s -> ms */ + int cr = mdev->sync_conf.rate; + + return d <= td ? cr : + d >= hd ? 0 : + cr + (cr * (td - d) / (hd - td)); +} + int w_make_resync_request(struct drbd_conf *mdev, struct drbd_work *w, int cancel) { @@ -446,7 +458,8 @@ int w_make_resync_request(struct drbd_conf *mdev, return 1; } - number = SLEEP_TIME * mdev->sync_conf.rate / ((BM_BLOCK_SIZE/1024)*HZ); + mdev->c_sync_rate = calc_resync_rate(mdev); + number = SLEEP_TIME * mdev->c_sync_rate / ((BM_BLOCK_SIZE / 1024) * HZ); pe = atomic_read(&mdev->rs_pending_cnt); mutex_lock(&mdev->data.mutex); -- cgit v1.1 From eedf386ae9d9e80a5669107e960090951e62f3a3 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Tue, 4 May 2010 16:31:03 +0200 Subject: drbd: Proc bits of new resync speed stuff Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_proc.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_proc.c b/drivers/block/drbd/drbd_proc.c index be3374b..81dea0a 100644 --- a/drivers/block/drbd/drbd_proc.c +++ b/drivers/block/drbd/drbd_proc.c @@ -73,14 +73,22 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) seq_printf(seq, "sync'ed:%3u.%u%% ", res / 10, res % 10); /* if more than 1 GB display in MB */ if (mdev->rs_total > 0x100000L) - seq_printf(seq, "(%lu/%lu)M\n\t", + seq_printf(seq, "(%lu/%lu)M", (unsigned long) Bit2KB(rs_left >> 10), (unsigned long) Bit2KB(mdev->rs_total >> 10)); else - seq_printf(seq, "(%lu/%lu)K\n\t", + seq_printf(seq, "(%lu/%lu)K", (unsigned long) Bit2KB(rs_left), (unsigned long) Bit2KB(mdev->rs_total)); + if (mdev->state.conn == C_SYNC_TARGET) + seq_printf(seq, " queue_delay: %d.%d ms\n\t", + mdev->data_delay / 1000, + (mdev->data_delay % 1000) / 100); + else if (mdev->state.conn == C_SYNC_SOURCE) + seq_printf(seq, " delay_probe: %d\n\t", + atomic_read(&mdev->delay_seq)); + /* see drivers/md/md.c * We do not want to overflow, so the order of operands and * the * 100 / 100 trick are important. We do a +1 to be @@ -128,6 +136,14 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) else seq_printf(seq, " (%ld)", dbdt); + if (mdev->state.conn == C_SYNC_TARGET) { + if (mdev->c_sync_rate > 1000) + seq_printf(seq, " want: %d,%03d", + mdev->c_sync_rate / 1000, mdev->c_sync_rate % 1000); + else + seq_printf(seq, " want: %d", mdev->c_sync_rate); + } + seq_printf(seq, " K/sec\n"); } -- cgit v1.1 From a8cdfd8d3bf0b6d2bbe792f5e74f54ccc6bc1d4f Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 5 May 2010 20:53:33 +0200 Subject: drbd: A fixes to the new resync speed code * Mention P_DELAY_PROBE in the packet naming array * Do not corrupt the mdev->data.work list in case the timer goes off before delay_probe_work got handled by the worker * Do not mod_timer() twice for a single delay_probe pair Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 1 + drivers/block/drbd/drbd_main.c | 9 +++++---- 2 files changed, 6 insertions(+), 4 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 210870e..37380d2 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -268,6 +268,7 @@ static inline const char *cmdname(enum drbd_packets cmd) [P_CSUM_RS_REQUEST] = "CsumRSRequest", [P_RS_IS_IN_SYNC] = "CsumRSIsInSync", [P_COMPRESSED_BITMAP] = "CBitmap", + [P_DELAY_PROBE] = "DelayProbe", [P_MAX_CMD] = NULL, }; diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 98785d0..44cc7b4 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2207,9 +2207,6 @@ static int drbd_send_delay_probe(struct drbd_conf *mdev, struct drbd_socket *ds) } mutex_unlock(&ds->mutex); - mdev->dp_volume_last = mdev->send_cnt; - mod_timer(&mdev->delay_probe_timer, jiffies + mdev->sync_conf.dp_interval * HZ / 10); - return ok; } @@ -2221,6 +2218,9 @@ static int drbd_send_delay_probes(struct drbd_conf *mdev) ok = drbd_send_delay_probe(mdev, &mdev->meta); ok = ok && drbd_send_delay_probe(mdev, &mdev->data); + mdev->dp_volume_last = mdev->send_cnt; + mod_timer(&mdev->delay_probe_timer, jiffies + mdev->sync_conf.dp_interval * HZ / 10); + return ok; } @@ -2374,7 +2374,8 @@ static void delay_probe_timer_fn(unsigned long data) { struct drbd_conf *mdev = (struct drbd_conf *) data; - drbd_queue_work(&mdev->data.work, &mdev->delay_probe_work); + if (list_empty(&mdev->delay_probe_work.list)) + drbd_queue_work(&mdev->data.work, &mdev->delay_probe_work); } /* Used to send write requests -- cgit v1.1 From 162f3ec7f026784ff2e216f19147d67e2f8ccd56 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 6 May 2010 15:19:30 +0200 Subject: drbd: Fixes to the new delay_probes code * Only send delay_probes with protocol 93 or newer * drbd_send_delay_probes() is called only from worker context, no atomic_t needed for delay_seq Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 6 +++--- drivers/block/drbd/drbd_main.c | 8 ++++---- drivers/block/drbd/drbd_proc.c | 3 +-- 3 files changed, 8 insertions(+), 9 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 37380d2..45d9a45 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -552,7 +552,7 @@ struct p_delay_probe { struct delay_probe { struct list_head list; - int seq_num; + unsigned int seq_num; struct timeval time; }; @@ -1048,9 +1048,9 @@ struct drbd_conf { char congestion_reason; /* Why we where congested... */ struct list_head delay_probes; /* protected by peer_seq_lock */ int data_delay; /* Delay of packets on the data-sock behind meta-sock */ - atomic_t delay_seq; /* To generate sequence numbers of delay probes */ + unsigned int delay_seq; /* To generate sequence numbers of delay probes */ struct timeval dps_time; /* delay-probes-start-time */ - int dp_volume_last; /* send_cnt of last delay probe */ + unsigned int dp_volume_last; /* send_cnt of last delay probe */ int c_sync_rate; /* current resync rate after delay_probe magic */ }; diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 44cc7b4..3aa0add 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2199,7 +2199,7 @@ static int drbd_send_delay_probe(struct drbd_conf *mdev, struct drbd_socket *ds) do_gettimeofday(&now); offset = now.tv_usec - mdev->dps_time.tv_usec + (now.tv_sec - mdev->dps_time.tv_sec) * 1000000; - dp.seq_num = cpu_to_be32(atomic_read(&mdev->delay_seq)); + dp.seq_num = cpu_to_be32(mdev->delay_seq); dp.offset = cpu_to_be32(offset); ok = _drbd_send_cmd(mdev, ds->socket, P_DELAY_PROBE, @@ -2213,7 +2213,8 @@ static int drbd_send_delay_probe(struct drbd_conf *mdev, struct drbd_socket *ds) static int drbd_send_delay_probes(struct drbd_conf *mdev) { int ok; - atomic_inc(&mdev->delay_seq); + + mdev->delay_seq++; do_gettimeofday(&mdev->dps_time); ok = drbd_send_delay_probe(mdev, &mdev->meta); ok = ok && drbd_send_delay_probe(mdev, &mdev->data); @@ -2355,7 +2356,7 @@ static int _drbd_send_zc_bio(struct drbd_conf *mdev, struct bio *bio) static void consider_delay_probes(struct drbd_conf *mdev) { - if (mdev->state.conn != C_SYNC_SOURCE) + if (mdev->state.conn != C_SYNC_SOURCE || mdev->agreed_pro_version < 93) return; if (mdev->dp_volume_last + mdev->sync_conf.dp_volume * 2 < mdev->send_cnt) @@ -2677,7 +2678,6 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) atomic_set(&mdev->net_cnt, 0); atomic_set(&mdev->packet_seq, 0); atomic_set(&mdev->pp_in_use, 0); - atomic_set(&mdev->delay_seq, 0); mutex_init(&mdev->md_io_mutex); mutex_init(&mdev->data.mutex); diff --git a/drivers/block/drbd/drbd_proc.c b/drivers/block/drbd/drbd_proc.c index 81dea0a..d0f1767 100644 --- a/drivers/block/drbd/drbd_proc.c +++ b/drivers/block/drbd/drbd_proc.c @@ -86,8 +86,7 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) mdev->data_delay / 1000, (mdev->data_delay % 1000) / 100); else if (mdev->state.conn == C_SYNC_SOURCE) - seq_printf(seq, " delay_probe: %d\n\t", - atomic_read(&mdev->delay_seq)); + seq_printf(seq, " delay_probe: %u\n\t", mdev->delay_seq); /* see drivers/md/md.c * We do not want to overflow, so the order of operands and -- cgit v1.1 From 708d740ed8242b84eefc63df144313a7308c7de5 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Mon, 3 May 2010 10:38:57 +0200 Subject: drbd: reduce sizeof struct drbd_epoch_entry by 8 byte by aligning members Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 45d9a45..1bc86dd 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -747,12 +747,8 @@ struct drbd_epoch_entry { struct hlist_node colision; sector_t sector; unsigned int size; - struct drbd_epoch *epoch; - - /* up to here, the struct layout is identical to drbd_request; - * we might be able to use that to our advantage... */ - unsigned int flags; + struct drbd_epoch *epoch; u64 block_id; }; -- cgit v1.1 From 45bb912bd5ea4d2b3a270a93cbdf767a0e2df6f5 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Fri, 14 May 2010 17:10:48 +0200 Subject: drbd: Allow drbd_epoch_entries to use multiple bios. This should allow for better performance if the lower level IO stack of the peers differs in limits exposed either via the queue, or via some merge_bvec_fn. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 90 +++++-- drivers/block/drbd/drbd_main.c | 19 +- drivers/block/drbd/drbd_nl.c | 18 +- drivers/block/drbd/drbd_receiver.c | 483 +++++++++++++++++++++---------------- drivers/block/drbd/drbd_worker.c | 178 ++++++++------ drivers/block/drbd/drbd_wrappers.h | 16 +- 6 files changed, 480 insertions(+), 324 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 1bc86dd..4b97f30 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -740,18 +740,6 @@ enum epoch_event { EV_CLEANUP = 32, /* used as flag */ }; -struct drbd_epoch_entry { - struct drbd_work w; - struct drbd_conf *mdev; - struct bio *private_bio; - struct hlist_node colision; - sector_t sector; - unsigned int size; - unsigned int flags; - struct drbd_epoch *epoch; - u64 block_id; -}; - struct drbd_wq_barrier { struct drbd_work w; struct completion done; @@ -762,17 +750,49 @@ struct digest_info { void *digest; }; -/* ee flag bits */ +struct drbd_epoch_entry { + struct drbd_work w; + struct hlist_node colision; + struct drbd_epoch *epoch; + struct drbd_conf *mdev; + struct page *pages; + atomic_t pending_bios; + unsigned int size; + /* see comments on ee flag bits below */ + unsigned long flags; + sector_t sector; + u64 block_id; +}; + +/* ee flag bits. + * While corresponding bios are in flight, the only modification will be + * set_bit WAS_ERROR, which has to be atomic. + * If no bios are in flight yet, or all have been completed, + * non-atomic modification to ee->flags is ok. + */ enum { __EE_CALL_AL_COMPLETE_IO, - __EE_CONFLICT_PENDING, __EE_MAY_SET_IN_SYNC, + + /* This epoch entry closes an epoch using a barrier. + * On sucessful completion, the epoch is released, + * and the P_BARRIER_ACK send. */ __EE_IS_BARRIER, + + /* In case a barrier failed, + * we need to resubmit without the barrier flag. */ + __EE_RESUBMITTED, + + /* we may have several bios per epoch entry. + * if any of those fail, we set this flag atomically + * from the endio callback */ + __EE_WAS_ERROR, }; #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) -#define EE_CONFLICT_PENDING (1<<__EE_CONFLICT_PENDING) #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) #define EE_IS_BARRIER (1<<__EE_IS_BARRIER) +#define EE_RESUBMITTED (1<<__EE_RESUBMITTED) +#define EE_WAS_ERROR (1<<__EE_WAS_ERROR) /* global flag bits */ enum { @@ -1441,7 +1461,8 @@ static inline void ov_oos_print(struct drbd_conf *mdev) } -extern void drbd_csum(struct drbd_conf *, struct crypto_hash *, struct bio *, void *); +extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *); +extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *); /* worker callbacks */ extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int); extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int); @@ -1465,6 +1486,8 @@ extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int); extern void resync_timer_fn(unsigned long data); /* drbd_receiver.c */ +extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, + const unsigned rw, const int fault_type); extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list); extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, u64 id, @@ -1620,6 +1643,41 @@ void drbd_bcast_ee(struct drbd_conf *mdev, * inline helper functions *************************/ +/* see also page_chain_add and friends in drbd_receiver.c */ +static inline struct page *page_chain_next(struct page *page) +{ + return (struct page *)page_private(page); +} +#define page_chain_for_each(page) \ + for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ + page = page_chain_next(page)) +#define page_chain_for_each_safe(page, n) \ + for (; page && ({ n = page_chain_next(page); 1; }); page = n) + +static inline int drbd_bio_has_active_page(struct bio *bio) +{ + struct bio_vec *bvec; + int i; + + __bio_for_each_segment(bvec, bio, i, 0) { + if (page_count(bvec->bv_page) > 1) + return 1; + } + + return 0; +} + +static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e) +{ + struct page *page = e->pages; + page_chain_for_each(page) { + if (page_count(page) > 1) + return 1; + } + return 0; +} + + static inline void drbd_state_lock(struct drbd_conf *mdev) { wait_event(mdev->misc_wait, diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 3aa0add..d0fabac 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2354,6 +2354,19 @@ static int _drbd_send_zc_bio(struct drbd_conf *mdev, struct bio *bio) return 1; } +static int _drbd_send_zc_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e) +{ + struct page *page = e->pages; + unsigned len = e->size; + page_chain_for_each(page) { + unsigned l = min_t(unsigned, len, PAGE_SIZE); + if (!_drbd_send_page(mdev, page, 0, l)) + return 0; + len -= l; + } + return 1; +} + static void consider_delay_probes(struct drbd_conf *mdev) { if (mdev->state.conn != C_SYNC_SOURCE || mdev->agreed_pro_version < 93) @@ -2430,7 +2443,7 @@ int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req) drbd_send(mdev, mdev->data.socket, &p, sizeof(p), MSG_MORE)); if (ok && dgs) { dgb = mdev->int_dig_out; - drbd_csum(mdev, mdev->integrity_w_tfm, req->master_bio, dgb); + drbd_csum_bio(mdev, mdev->integrity_w_tfm, req->master_bio, dgb); ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, MSG_MORE); } if (ok) { @@ -2483,11 +2496,11 @@ int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, sizeof(p), MSG_MORE); if (ok && dgs) { dgb = mdev->int_dig_out; - drbd_csum(mdev, mdev->integrity_w_tfm, e->private_bio, dgb); + drbd_csum_ee(mdev, mdev->integrity_w_tfm, e, dgb); ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, MSG_MORE); } if (ok) - ok = _drbd_send_zc_bio(mdev, e->private_bio); + ok = _drbd_send_zc_ee(mdev, e); drbd_put_data_sock(mdev); diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 93d1506..28ef76b 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -2215,9 +2215,9 @@ void drbd_bcast_ee(struct drbd_conf *mdev, { struct cn_msg *cn_reply; struct drbd_nl_cfg_reply *reply; - struct bio_vec *bvec; unsigned short *tl; - int i; + struct page *page; + unsigned len; if (!e) return; @@ -2255,11 +2255,15 @@ void drbd_bcast_ee(struct drbd_conf *mdev, put_unaligned(T_ee_data, tl++); put_unaligned(e->size, tl++); - __bio_for_each_segment(bvec, e->private_bio, i, 0) { - void *d = kmap(bvec->bv_page); - memcpy(tl, d + bvec->bv_offset, bvec->bv_len); - kunmap(bvec->bv_page); - tl=(unsigned short*)((char*)tl + bvec->bv_len); + len = e->size; + page = e->pages; + page_chain_for_each(page) { + void *d = kmap_atomic(page, KM_USER0); + unsigned l = min_t(unsigned, len, PAGE_SIZE); + memcpy(tl, d, l); + kunmap_atomic(d, KM_USER0); + tl = (unsigned short*)((char*)tl + l); + len -= l; } put_unaligned(TT_END, tl++); /* Close the tag list */ diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index fee0d24..388a3e8 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -80,30 +80,124 @@ static struct drbd_epoch *previous_epoch(struct drbd_conf *mdev, struct drbd_epo #define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN) -static struct page *drbd_pp_first_page_or_try_alloc(struct drbd_conf *mdev) +/* + * some helper functions to deal with single linked page lists, + * page->private being our "next" pointer. + */ + +/* If at least n pages are linked at head, get n pages off. + * Otherwise, don't modify head, and return NULL. + * Locking is the responsibility of the caller. + */ +static struct page *page_chain_del(struct page **head, int n) +{ + struct page *page; + struct page *tmp; + + BUG_ON(!n); + BUG_ON(!head); + + page = *head; + while (page) { + tmp = page_chain_next(page); + if (--n == 0) + break; /* found sufficient pages */ + if (tmp == NULL) + /* insufficient pages, don't use any of them. */ + return NULL; + page = tmp; + } + + /* add end of list marker for the returned list */ + set_page_private(page, 0); + /* actual return value, and adjustment of head */ + page = *head; + *head = tmp; + return page; +} + +/* may be used outside of locks to find the tail of a (usually short) + * "private" page chain, before adding it back to a global chain head + * with page_chain_add() under a spinlock. */ +static struct page *page_chain_tail(struct page *page, int *len) +{ + struct page *tmp; + int i = 1; + while ((tmp = page_chain_next(page))) + ++i, page = tmp; + if (len) + *len = i; + return page; +} + +static int page_chain_free(struct page *page) +{ + struct page *tmp; + int i = 0; + page_chain_for_each_safe(page, tmp) { + put_page(page); + ++i; + } + return i; +} + +static void page_chain_add(struct page **head, + struct page *chain_first, struct page *chain_last) +{ +#if 1 + struct page *tmp; + tmp = page_chain_tail(chain_first, NULL); + BUG_ON(tmp != chain_last); +#endif + + /* add chain to head */ + set_page_private(chain_last, (unsigned long)*head); + *head = chain_first; +} + +static struct page *drbd_pp_first_pages_or_try_alloc(struct drbd_conf *mdev, int number) { struct page *page = NULL; + struct page *tmp = NULL; + int i = 0; /* Yes, testing drbd_pp_vacant outside the lock is racy. * So what. It saves a spin_lock. */ - if (drbd_pp_vacant > 0) { + if (drbd_pp_vacant >= number) { spin_lock(&drbd_pp_lock); - page = drbd_pp_pool; - if (page) { - drbd_pp_pool = (struct page *)page_private(page); - set_page_private(page, 0); /* just to be polite */ - drbd_pp_vacant--; - } + page = page_chain_del(&drbd_pp_pool, number); + if (page) + drbd_pp_vacant -= number; spin_unlock(&drbd_pp_lock); + if (page) + return page; } + /* GFP_TRY, because we must not cause arbitrary write-out: in a DRBD * "criss-cross" setup, that might cause write-out on some other DRBD, * which in turn might block on the other node at this very place. */ - if (!page) - page = alloc_page(GFP_TRY); - if (page) - atomic_inc(&mdev->pp_in_use); - return page; + for (i = 0; i < number; i++) { + tmp = alloc_page(GFP_TRY); + if (!tmp) + break; + set_page_private(tmp, (unsigned long)page); + page = tmp; + } + + if (i == number) + return page; + + /* Not enough pages immediately available this time. + * No need to jump around here, drbd_pp_alloc will retry this + * function "soon". */ + if (page) { + tmp = page_chain_tail(page, NULL); + spin_lock(&drbd_pp_lock); + page_chain_add(&drbd_pp_pool, page, tmp); + drbd_pp_vacant += i; + spin_unlock(&drbd_pp_lock); + } + return NULL; } /* kick lower level device, if we have more than (arbitrary number) @@ -127,7 +221,7 @@ static void reclaim_net_ee(struct drbd_conf *mdev, struct list_head *to_be_freed list_for_each_safe(le, tle, &mdev->net_ee) { e = list_entry(le, struct drbd_epoch_entry, w.list); - if (drbd_bio_has_active_page(e->private_bio)) + if (drbd_ee_has_active_page(e)) break; list_move(le, to_be_freed); } @@ -148,32 +242,34 @@ static void drbd_kick_lo_and_reclaim_net(struct drbd_conf *mdev) } /** - * drbd_pp_alloc() - Returns a page, fails only if a signal comes in + * drbd_pp_alloc() - Returns @number pages, retries forever (or until signalled) * @mdev: DRBD device. - * @retry: whether or not to retry allocation forever (or until signalled) + * @number: number of pages requested + * @retry: whether to retry, if not enough pages are available right now + * + * Tries to allocate number pages, first from our own page pool, then from + * the kernel, unless this allocation would exceed the max_buffers setting. + * Possibly retry until DRBD frees sufficient pages somewhere else. * - * Tries to allocate a page, first from our own page pool, then from the - * kernel, unless this allocation would exceed the max_buffers setting. - * If @retry is non-zero, retry until DRBD frees a page somewhere else. + * Returns a page chain linked via page->private. */ -static struct page *drbd_pp_alloc(struct drbd_conf *mdev, int retry) +static struct page *drbd_pp_alloc(struct drbd_conf *mdev, unsigned number, bool retry) { struct page *page = NULL; DEFINE_WAIT(wait); - if (atomic_read(&mdev->pp_in_use) < mdev->net_conf->max_buffers) { - page = drbd_pp_first_page_or_try_alloc(mdev); - if (page) - return page; - } + /* Yes, we may run up to @number over max_buffers. If we + * follow it strictly, the admin will get it wrong anyways. */ + if (atomic_read(&mdev->pp_in_use) < mdev->net_conf->max_buffers) + page = drbd_pp_first_pages_or_try_alloc(mdev, number); - for (;;) { + while (page == NULL) { prepare_to_wait(&drbd_pp_wait, &wait, TASK_INTERRUPTIBLE); drbd_kick_lo_and_reclaim_net(mdev); if (atomic_read(&mdev->pp_in_use) < mdev->net_conf->max_buffers) { - page = drbd_pp_first_page_or_try_alloc(mdev); + page = drbd_pp_first_pages_or_try_alloc(mdev, number); if (page) break; } @@ -190,62 +286,32 @@ static struct page *drbd_pp_alloc(struct drbd_conf *mdev, int retry) } finish_wait(&drbd_pp_wait, &wait); + if (page) + atomic_add(number, &mdev->pp_in_use); return page; } /* Must not be used from irq, as that may deadlock: see drbd_pp_alloc. - * Is also used from inside an other spin_lock_irq(&mdev->req_lock) */ + * Is also used from inside an other spin_lock_irq(&mdev->req_lock); + * Either links the page chain back to the global pool, + * or returns all pages to the system. */ static void drbd_pp_free(struct drbd_conf *mdev, struct page *page) { - int free_it; - - spin_lock(&drbd_pp_lock); - if (drbd_pp_vacant > (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE)*minor_count) { - free_it = 1; - } else { - set_page_private(page, (unsigned long)drbd_pp_pool); - drbd_pp_pool = page; - drbd_pp_vacant++; - free_it = 0; - } - spin_unlock(&drbd_pp_lock); - - atomic_dec(&mdev->pp_in_use); - - if (free_it) - __free_page(page); - - wake_up(&drbd_pp_wait); -} - -static void drbd_pp_free_bio_pages(struct drbd_conf *mdev, struct bio *bio) -{ - struct page *p_to_be_freed = NULL; - struct page *page; - struct bio_vec *bvec; int i; - - spin_lock(&drbd_pp_lock); - __bio_for_each_segment(bvec, bio, i, 0) { - if (drbd_pp_vacant > (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE)*minor_count) { - set_page_private(bvec->bv_page, (unsigned long)p_to_be_freed); - p_to_be_freed = bvec->bv_page; - } else { - set_page_private(bvec->bv_page, (unsigned long)drbd_pp_pool); - drbd_pp_pool = bvec->bv_page; - drbd_pp_vacant++; - } - } - spin_unlock(&drbd_pp_lock); - atomic_sub(bio->bi_vcnt, &mdev->pp_in_use); - - while (p_to_be_freed) { - page = p_to_be_freed; - p_to_be_freed = (struct page *)page_private(page); - set_page_private(page, 0); /* just to be polite */ - put_page(page); + if (drbd_pp_vacant > (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE)*minor_count) + i = page_chain_free(page); + else { + struct page *tmp; + tmp = page_chain_tail(page, &i); + spin_lock(&drbd_pp_lock); + page_chain_add(&drbd_pp_pool, page, tmp); + drbd_pp_vacant += i; + spin_unlock(&drbd_pp_lock); } - + atomic_sub(i, &mdev->pp_in_use); + i = atomic_read(&mdev->pp_in_use); + if (i < 0) + dev_warn(DEV, "ASSERTION FAILED: pp_in_use: %d < 0\n", i); wake_up(&drbd_pp_wait); } @@ -270,11 +336,9 @@ struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, unsigned int data_size, gfp_t gfp_mask) __must_hold(local) { - struct request_queue *q; struct drbd_epoch_entry *e; struct page *page; - struct bio *bio; - unsigned int ds; + unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT; if (FAULT_ACTIVE(mdev, DRBD_FAULT_AL_EE)) return NULL; @@ -286,84 +350,32 @@ struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, return NULL; } - bio = bio_alloc(gfp_mask & ~__GFP_HIGHMEM, div_ceil(data_size, PAGE_SIZE)); - if (!bio) { - if (!(gfp_mask & __GFP_NOWARN)) - dev_err(DEV, "alloc_ee: Allocation of a bio failed\n"); - goto fail1; - } - - bio->bi_bdev = mdev->ldev->backing_bdev; - bio->bi_sector = sector; - - ds = data_size; - while (ds) { - page = drbd_pp_alloc(mdev, (gfp_mask & __GFP_WAIT)); - if (!page) { - if (!(gfp_mask & __GFP_NOWARN)) - dev_err(DEV, "alloc_ee: Allocation of a page failed\n"); - goto fail2; - } - if (!bio_add_page(bio, page, min_t(int, ds, PAGE_SIZE), 0)) { - drbd_pp_free(mdev, page); - dev_err(DEV, "alloc_ee: bio_add_page(s=%llu," - "data_size=%u,ds=%u) failed\n", - (unsigned long long)sector, data_size, ds); - - q = bdev_get_queue(bio->bi_bdev); - if (q->merge_bvec_fn) { - struct bvec_merge_data bvm = { - .bi_bdev = bio->bi_bdev, - .bi_sector = bio->bi_sector, - .bi_size = bio->bi_size, - .bi_rw = bio->bi_rw, - }; - int l = q->merge_bvec_fn(q, &bvm, - &bio->bi_io_vec[bio->bi_vcnt]); - dev_err(DEV, "merge_bvec_fn() = %d\n", l); - } - - /* dump more of the bio. */ - dev_err(DEV, "bio->bi_max_vecs = %d\n", bio->bi_max_vecs); - dev_err(DEV, "bio->bi_vcnt = %d\n", bio->bi_vcnt); - dev_err(DEV, "bio->bi_size = %d\n", bio->bi_size); - dev_err(DEV, "bio->bi_phys_segments = %d\n", bio->bi_phys_segments); - - goto fail2; - break; - } - ds -= min_t(int, ds, PAGE_SIZE); - } - - D_ASSERT(data_size == bio->bi_size); - - bio->bi_private = e; - e->mdev = mdev; - e->sector = sector; - e->size = bio->bi_size; + page = drbd_pp_alloc(mdev, nr_pages, (gfp_mask & __GFP_WAIT)); + if (!page) + goto fail; - e->private_bio = bio; - e->block_id = id; INIT_HLIST_NODE(&e->colision); e->epoch = NULL; + e->mdev = mdev; + e->pages = page; + atomic_set(&e->pending_bios, 0); + e->size = data_size; e->flags = 0; + e->sector = sector; + e->sector = sector; + e->block_id = id; return e; - fail2: - drbd_pp_free_bio_pages(mdev, bio); - bio_put(bio); - fail1: + fail: mempool_free(e, drbd_ee_mempool); - return NULL; } void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e) { - struct bio *bio = e->private_bio; - drbd_pp_free_bio_pages(mdev, bio); - bio_put(bio); + drbd_pp_free(mdev, e->pages); + D_ASSERT(atomic_read(&e->pending_bios) == 0); D_ASSERT(hlist_unhashed(&e->colision)); mempool_free(e, drbd_ee_mempool); } @@ -1121,6 +1133,90 @@ void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo) } /** + * drbd_submit_ee() + * @mdev: DRBD device. + * @e: epoch entry + * @rw: flag field, see bio->bi_rw + */ +/* TODO allocate from our own bio_set. */ +int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, + const unsigned rw, const int fault_type) +{ + struct bio *bios = NULL; + struct bio *bio; + struct page *page = e->pages; + sector_t sector = e->sector; + unsigned ds = e->size; + unsigned n_bios = 0; + unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT; + + /* In most cases, we will only need one bio. But in case the lower + * level restrictions happen to be different at this offset on this + * side than those of the sending peer, we may need to submit the + * request in more than one bio. */ +next_bio: + bio = bio_alloc(GFP_NOIO, nr_pages); + if (!bio) { + dev_err(DEV, "submit_ee: Allocation of a bio failed\n"); + goto fail; + } + /* > e->sector, unless this is the first bio */ + bio->bi_sector = sector; + bio->bi_bdev = mdev->ldev->backing_bdev; + /* we special case some flags in the multi-bio case, see below + * (BIO_RW_UNPLUG, BIO_RW_BARRIER) */ + bio->bi_rw = rw; + bio->bi_private = e; + bio->bi_end_io = drbd_endio_sec; + + bio->bi_next = bios; + bios = bio; + ++n_bios; + + page_chain_for_each(page) { + unsigned len = min_t(unsigned, ds, PAGE_SIZE); + if (!bio_add_page(bio, page, len, 0)) { + /* a single page must always be possible! */ + BUG_ON(bio->bi_vcnt == 0); + goto next_bio; + } + ds -= len; + sector += len >> 9; + --nr_pages; + } + D_ASSERT(page == NULL); + D_ASSERT(ds == 0); + + atomic_set(&e->pending_bios, n_bios); + do { + bio = bios; + bios = bios->bi_next; + bio->bi_next = NULL; + + /* strip off BIO_RW_UNPLUG unless it is the last bio */ + if (bios) + bio->bi_rw &= ~(1<bi_next) + bios->bi_rw &= ~(1<bi_next; + bio_put(bio); + } + return -ENOMEM; +} + +/** * w_e_reissue() - Worker callback; Resubmit a bio, without BIO_RW_BARRIER set * @mdev: DRBD device. * @w: work object. @@ -1129,8 +1225,6 @@ void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo) int w_e_reissue(struct drbd_conf *mdev, struct drbd_work *w, int cancel) __releases(local) { struct drbd_epoch_entry *e = (struct drbd_epoch_entry *)w; - struct bio *bio = e->private_bio; - /* We leave DE_CONTAINS_A_BARRIER and EE_IS_BARRIER in place, (and DE_BARRIER_IN_NEXT_EPOCH_ISSUED in the previous Epoch) so that we can finish that epoch in drbd_may_finish_epoch(). @@ -1144,33 +1238,17 @@ int w_e_reissue(struct drbd_conf *mdev, struct drbd_work *w, int cancel) __relea if (previous_epoch(mdev, e->epoch)) dev_warn(DEV, "Write ordering was not enforced (one time event)\n"); - /* prepare bio for re-submit, - * re-init volatile members */ /* we still have a local reference, * get_ldev was done in receive_Data. */ - bio->bi_bdev = mdev->ldev->backing_bdev; - bio->bi_sector = e->sector; - bio->bi_size = e->size; - bio->bi_idx = 0; - - bio->bi_flags &= ~(BIO_POOL_MASK - 1); - bio->bi_flags |= 1 << BIO_UPTODATE; - - /* don't know whether this is necessary: */ - bio->bi_phys_segments = 0; - bio->bi_next = NULL; - - /* these should be unchanged: */ - /* bio->bi_end_io = drbd_endio_write_sec; */ - /* bio->bi_vcnt = whatever; */ e->w.cb = e_end_block; - - /* This is no longer a barrier request. */ - bio->bi_rw &= ~(1UL << BIO_RW_BARRIER); - - drbd_generic_make_request(mdev, DRBD_FAULT_DT_WR, bio); - + if (drbd_submit_ee(mdev, e, WRITE, DRBD_FAULT_DT_WR) != 0) { + /* drbd_submit_ee fails for one reason only: + * if was not able to allocate sufficient bios. + * requeue, try again later. */ + e->w.cb = w_e_reissue; + drbd_queue_work(&mdev->data.work, &e->w); + } return 1; } @@ -1264,10 +1342,8 @@ read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __ { const sector_t capacity = drbd_get_capacity(mdev->this_bdev); struct drbd_epoch_entry *e; - struct bio_vec *bvec; struct page *page; - struct bio *bio; - int dgs, ds, i, rr; + int dgs, ds, rr; void *dig_in = mdev->int_dig_in; void *dig_vv = mdev->int_dig_vv; unsigned long *data; @@ -1304,28 +1380,29 @@ read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __ e = drbd_alloc_ee(mdev, id, sector, data_size, GFP_NOIO); if (!e) return NULL; - bio = e->private_bio; + ds = data_size; - bio_for_each_segment(bvec, bio, i) { - page = bvec->bv_page; + page = e->pages; + page_chain_for_each(page) { + unsigned len = min_t(int, ds, PAGE_SIZE); data = kmap(page); - rr = drbd_recv(mdev, data, min_t(int, ds, PAGE_SIZE)); + rr = drbd_recv(mdev, data, len); if (FAULT_ACTIVE(mdev, DRBD_FAULT_RECEIVE)) { dev_err(DEV, "Fault injection: Corrupting data on receive\n"); data[0] = data[0] ^ (unsigned long)-1; } kunmap(page); - if (rr != min_t(int, ds, PAGE_SIZE)) { + if (rr != len) { drbd_free_ee(mdev, e); dev_warn(DEV, "short read receiving data: read %d expected %d\n", - rr, min_t(int, ds, PAGE_SIZE)); + rr, len); return NULL; } ds -= rr; } if (dgs) { - drbd_csum(mdev, mdev->integrity_r_tfm, bio, dig_vv); + drbd_csum_ee(mdev, mdev->integrity_r_tfm, e, dig_vv); if (memcmp(dig_in, dig_vv, dgs)) { dev_err(DEV, "Digest integrity check FAILED.\n"); drbd_bcast_ee(mdev, "digest failed", @@ -1350,7 +1427,7 @@ static int drbd_drain_block(struct drbd_conf *mdev, int data_size) if (!data_size) return TRUE; - page = drbd_pp_alloc(mdev, 1); + page = drbd_pp_alloc(mdev, 1, 1); data = kmap(page); while (data_size) { @@ -1414,7 +1491,7 @@ static int recv_dless_read(struct drbd_conf *mdev, struct drbd_request *req, } if (dgs) { - drbd_csum(mdev, mdev->integrity_r_tfm, bio, dig_vv); + drbd_csum_bio(mdev, mdev->integrity_r_tfm, bio, dig_vv); if (memcmp(dig_in, dig_vv, dgs)) { dev_err(DEV, "Digest integrity check FAILED. Broken NICs?\n"); return 0; @@ -1435,7 +1512,7 @@ static int e_end_resync_block(struct drbd_conf *mdev, struct drbd_work *w, int u D_ASSERT(hlist_unhashed(&e->colision)); - if (likely(drbd_bio_uptodate(e->private_bio))) { + if (likely((e->flags & EE_WAS_ERROR) == 0)) { drbd_set_in_sync(mdev, sector, e->size); ok = drbd_send_ack(mdev, P_RS_WRITE_ACK, e); } else { @@ -1454,30 +1531,28 @@ static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_si struct drbd_epoch_entry *e; e = read_in_block(mdev, ID_SYNCER, sector, data_size); - if (!e) { - put_ldev(mdev); - return FALSE; - } + if (!e) + goto fail; dec_rs_pending(mdev); - e->private_bio->bi_end_io = drbd_endio_write_sec; - e->private_bio->bi_rw = WRITE; - e->w.cb = e_end_resync_block; - inc_unacked(mdev); /* corresponding dec_unacked() in e_end_resync_block() * respective _drbd_clear_done_ee */ + e->w.cb = e_end_resync_block; + spin_lock_irq(&mdev->req_lock); list_add(&e->w.list, &mdev->sync_ee); spin_unlock_irq(&mdev->req_lock); - drbd_generic_make_request(mdev, DRBD_FAULT_RS_WR, e->private_bio); - /* accounting done in endio */ + if (drbd_submit_ee(mdev, e, WRITE, DRBD_FAULT_RS_WR) == 0) + return TRUE; - maybe_kick_lo(mdev); - return TRUE; + drbd_free_ee(mdev, e); +fail: + put_ldev(mdev); + return FALSE; } static int receive_DataReply(struct drbd_conf *mdev, struct p_header *h) @@ -1572,7 +1647,7 @@ static int e_end_block(struct drbd_conf *mdev, struct drbd_work *w, int cancel) } if (mdev->net_conf->wire_protocol == DRBD_PROT_C) { - if (likely(drbd_bio_uptodate(e->private_bio))) { + if (likely((e->flags & EE_WAS_ERROR) == 0)) { pcmd = (mdev->state.conn >= C_SYNC_SOURCE && mdev->state.conn <= C_PAUSED_SYNC_T && e->flags & EE_MAY_SET_IN_SYNC) ? @@ -1718,7 +1793,6 @@ static int receive_Data(struct drbd_conf *mdev, struct p_header *h) return FALSE; } - e->private_bio->bi_end_io = drbd_endio_write_sec; e->w.cb = e_end_block; spin_lock(&mdev->epoch_lock); @@ -1914,12 +1988,8 @@ static int receive_Data(struct drbd_conf *mdev, struct p_header *h) drbd_al_begin_io(mdev, e->sector); } - e->private_bio->bi_rw = rw; - drbd_generic_make_request(mdev, DRBD_FAULT_DT_WR, e->private_bio); - /* accounting done in endio */ - - maybe_kick_lo(mdev); - return TRUE; + if (drbd_submit_ee(mdev, e, rw, DRBD_FAULT_DT_WR) == 0) + return TRUE; out_interrupted: /* yes, the epoch_size now is imbalanced. @@ -1977,9 +2047,6 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) return FALSE; } - e->private_bio->bi_rw = READ; - e->private_bio->bi_end_io = drbd_endio_read_sec; - switch (h->command) { case P_DATA_REQUEST: e->w.cb = w_e_end_data_req; @@ -2073,10 +2140,8 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) inc_unacked(mdev); - drbd_generic_make_request(mdev, fault_type, e->private_bio); - maybe_kick_lo(mdev); - - return TRUE; + if (drbd_submit_ee(mdev, e, READ, fault_type) == 0) + return TRUE; out_free_e: kfree(di); @@ -3837,7 +3902,7 @@ static void drbd_disconnect(struct drbd_conf *mdev) dev_info(DEV, "net_ee not empty, killed %u entries\n", i); i = atomic_read(&mdev->pp_in_use); if (i) - dev_info(DEV, "pp_in_use = %u, expected 0\n", i); + dev_info(DEV, "pp_in_use = %d, expected 0\n", i); D_ASSERT(list_empty(&mdev->read_ee)); D_ASSERT(list_empty(&mdev->active_ee)); diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index 0bbecf4..d771b1e 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -47,8 +47,7 @@ static int w_make_ov_request(struct drbd_conf *mdev, struct drbd_work *w, int ca /* defined here: drbd_md_io_complete - drbd_endio_write_sec - drbd_endio_read_sec + drbd_endio_sec drbd_endio_pri * more endio handlers: @@ -85,27 +84,10 @@ void drbd_md_io_complete(struct bio *bio, int error) /* reads on behalf of the partner, * "submitted" by the receiver */ -void drbd_endio_read_sec(struct bio *bio, int error) __releases(local) +void drbd_endio_read_sec_final(struct drbd_epoch_entry *e) __releases(local) { unsigned long flags = 0; - struct drbd_epoch_entry *e = NULL; - struct drbd_conf *mdev; - int uptodate = bio_flagged(bio, BIO_UPTODATE); - - e = bio->bi_private; - mdev = e->mdev; - - if (error) - dev_warn(DEV, "read: error=%d s=%llus\n", error, - (unsigned long long)e->sector); - if (!error && !uptodate) { - dev_warn(DEV, "read: setting error to -EIO s=%llus\n", - (unsigned long long)e->sector); - /* strange behavior of some lower level drivers... - * fail the request by clearing the uptodate flag, - * but do not return any error?! */ - error = -EIO; - } + struct drbd_conf *mdev = e->mdev; D_ASSERT(e->block_id != ID_VACANT); @@ -114,49 +96,38 @@ void drbd_endio_read_sec(struct bio *bio, int error) __releases(local) list_del(&e->w.list); if (list_empty(&mdev->read_ee)) wake_up(&mdev->ee_wait); + if (test_bit(__EE_WAS_ERROR, &e->flags)) + __drbd_chk_io_error(mdev, FALSE); spin_unlock_irqrestore(&mdev->req_lock, flags); - drbd_chk_io_error(mdev, error, FALSE); drbd_queue_work(&mdev->data.work, &e->w); put_ldev(mdev); } +static int is_failed_barrier(int ee_flags) +{ + return (ee_flags & (EE_IS_BARRIER|EE_WAS_ERROR|EE_RESUBMITTED)) + == (EE_IS_BARRIER|EE_WAS_ERROR); +} + /* writes on behalf of the partner, or resync writes, - * "submitted" by the receiver. - */ -void drbd_endio_write_sec(struct bio *bio, int error) __releases(local) + * "submitted" by the receiver, final stage. */ +static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(local) { unsigned long flags = 0; - struct drbd_epoch_entry *e = NULL; - struct drbd_conf *mdev; + struct drbd_conf *mdev = e->mdev; sector_t e_sector; int do_wake; int is_syncer_req; int do_al_complete_io; - int uptodate = bio_flagged(bio, BIO_UPTODATE); - int is_barrier = bio_rw_flagged(bio, BIO_RW_BARRIER); - e = bio->bi_private; - mdev = e->mdev; - - if (error) - dev_warn(DEV, "write: error=%d s=%llus\n", error, - (unsigned long long)e->sector); - if (!error && !uptodate) { - dev_warn(DEV, "write: setting error to -EIO s=%llus\n", - (unsigned long long)e->sector); - /* strange behavior of some lower level drivers... - * fail the request by clearing the uptodate flag, - * but do not return any error?! */ - error = -EIO; - } - - /* error == -ENOTSUPP would be a better test, - * alas it is not reliable */ - if (error && is_barrier && e->flags & EE_IS_BARRIER) { + /* if this is a failed barrier request, disable use of barriers, + * and schedule for resubmission */ + if (is_failed_barrier(e->flags)) { drbd_bump_write_ordering(mdev, WO_bdev_flush); spin_lock_irqsave(&mdev->req_lock, flags); list_del(&e->w.list); + e->flags |= EE_RESUBMITTED; e->w.cb = w_e_reissue; /* put_ldev actually happens below, once we come here again. */ __release(local); @@ -167,17 +138,16 @@ void drbd_endio_write_sec(struct bio *bio, int error) __releases(local) D_ASSERT(e->block_id != ID_VACANT); - spin_lock_irqsave(&mdev->req_lock, flags); - mdev->writ_cnt += e->size >> 9; - is_syncer_req = is_syncer_block_id(e->block_id); - /* after we moved e to done_ee, * we may no longer access it, * it may be freed/reused already! * (as soon as we release the req_lock) */ e_sector = e->sector; do_al_complete_io = e->flags & EE_CALL_AL_COMPLETE_IO; + is_syncer_req = is_syncer_block_id(e->block_id); + spin_lock_irqsave(&mdev->req_lock, flags); + mdev->writ_cnt += e->size >> 9; list_del(&e->w.list); /* has been on active_ee or sync_ee */ list_add_tail(&e->w.list, &mdev->done_ee); @@ -190,7 +160,7 @@ void drbd_endio_write_sec(struct bio *bio, int error) __releases(local) ? list_empty(&mdev->sync_ee) : list_empty(&mdev->active_ee); - if (error) + if (test_bit(__EE_WAS_ERROR, &e->flags)) __drbd_chk_io_error(mdev, FALSE); spin_unlock_irqrestore(&mdev->req_lock, flags); @@ -205,7 +175,42 @@ void drbd_endio_write_sec(struct bio *bio, int error) __releases(local) wake_asender(mdev); put_ldev(mdev); +} +/* writes on behalf of the partner, or resync writes, + * "submitted" by the receiver. + */ +void drbd_endio_sec(struct bio *bio, int error) +{ + struct drbd_epoch_entry *e = bio->bi_private; + struct drbd_conf *mdev = e->mdev; + int uptodate = bio_flagged(bio, BIO_UPTODATE); + int is_write = bio_data_dir(bio) == WRITE; + + if (error) + dev_warn(DEV, "%s: error=%d s=%llus\n", + is_write ? "write" : "read", error, + (unsigned long long)e->sector); + if (!error && !uptodate) { + dev_warn(DEV, "%s: setting error to -EIO s=%llus\n", + is_write ? "write" : "read", + (unsigned long long)e->sector); + /* strange behavior of some lower level drivers... + * fail the request by clearing the uptodate flag, + * but do not return any error?! */ + error = -EIO; + } + + if (error) + set_bit(__EE_WAS_ERROR, &e->flags); + + bio_put(bio); /* no need for the bio anymore */ + if (atomic_dec_and_test(&e->pending_bios)) { + if (is_write) + drbd_endio_write_sec_final(e); + else + drbd_endio_read_sec_final(e); + } } /* read, readA or write requests on R_PRIMARY coming from drbd_make_request @@ -295,7 +300,34 @@ int w_resync_inactive(struct drbd_conf *mdev, struct drbd_work *w, int cancel) return 1; /* Simply ignore this! */ } -void drbd_csum(struct drbd_conf *mdev, struct crypto_hash *tfm, struct bio *bio, void *digest) +void drbd_csum_ee(struct drbd_conf *mdev, struct crypto_hash *tfm, struct drbd_epoch_entry *e, void *digest) +{ + struct hash_desc desc; + struct scatterlist sg; + struct page *page = e->pages; + struct page *tmp; + unsigned len; + + desc.tfm = tfm; + desc.flags = 0; + + sg_init_table(&sg, 1); + crypto_hash_init(&desc); + + while ((tmp = page_chain_next(page))) { + /* all but the last page will be fully used */ + sg_set_page(&sg, page, PAGE_SIZE, 0); + crypto_hash_update(&desc, &sg, sg.length); + page = tmp; + } + /* and now the last, possibly only partially used page */ + len = e->size & (PAGE_SIZE - 1); + sg_set_page(&sg, page, len ?: PAGE_SIZE, 0); + crypto_hash_update(&desc, &sg, sg.length); + crypto_hash_final(&desc, digest); +} + +void drbd_csum_bio(struct drbd_conf *mdev, struct crypto_hash *tfm, struct bio *bio, void *digest) { struct hash_desc desc; struct scatterlist sg; @@ -329,11 +361,11 @@ static int w_e_send_csum(struct drbd_conf *mdev, struct drbd_work *w, int cancel return 1; } - if (likely(drbd_bio_uptodate(e->private_bio))) { + if (likely((e->flags & EE_WAS_ERROR) == 0)) { digest_size = crypto_hash_digestsize(mdev->csums_tfm); digest = kmalloc(digest_size, GFP_NOIO); if (digest) { - drbd_csum(mdev, mdev->csums_tfm, e->private_bio, digest); + drbd_csum_ee(mdev, mdev->csums_tfm, e, digest); inc_rs_pending(mdev); ok = drbd_send_drequest_csum(mdev, @@ -369,23 +401,21 @@ static int read_for_csum(struct drbd_conf *mdev, sector_t sector, int size) /* GFP_TRY, because if there is no memory available right now, this may * be rescheduled for later. It is "only" background resync, after all. */ e = drbd_alloc_ee(mdev, DRBD_MAGIC+0xbeef, sector, size, GFP_TRY); - if (!e) { - put_ldev(mdev); - return 2; - } + if (!e) + goto fail; spin_lock_irq(&mdev->req_lock); list_add(&e->w.list, &mdev->read_ee); spin_unlock_irq(&mdev->req_lock); - e->private_bio->bi_end_io = drbd_endio_read_sec; - e->private_bio->bi_rw = READ; e->w.cb = w_e_send_csum; + if (drbd_submit_ee(mdev, e, READ, DRBD_FAULT_RS_RD) == 0) + return 1; - mdev->read_cnt += size >> 9; - drbd_generic_make_request(mdev, DRBD_FAULT_RS_RD, e->private_bio); - - return 1; + drbd_free_ee(mdev, e); +fail: + put_ldev(mdev); + return 2; } void resync_timer_fn(unsigned long data) @@ -819,7 +849,7 @@ out: /* helper */ static void move_to_net_ee_or_free(struct drbd_conf *mdev, struct drbd_epoch_entry *e) { - if (drbd_bio_has_active_page(e->private_bio)) { + if (drbd_ee_has_active_page(e)) { /* This might happen if sendpage() has not finished */ spin_lock_irq(&mdev->req_lock); list_add_tail(&e->w.list, &mdev->net_ee); @@ -845,7 +875,7 @@ int w_e_end_data_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) return 1; } - if (likely(drbd_bio_uptodate(e->private_bio))) { + if (likely((e->flags & EE_WAS_ERROR) == 0)) { ok = drbd_send_block(mdev, P_DATA_REPLY, e); } else { if (__ratelimit(&drbd_ratelimit_state)) @@ -886,7 +916,7 @@ int w_e_end_rsdata_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) put_ldev(mdev); } - if (likely(drbd_bio_uptodate(e->private_bio))) { + if (likely((e->flags & EE_WAS_ERROR) == 0)) { if (likely(mdev->state.pdsk >= D_INCONSISTENT)) { inc_rs_pending(mdev); ok = drbd_send_block(mdev, P_RS_DATA_REPLY, e); @@ -934,7 +964,7 @@ int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) di = (struct digest_info *)(unsigned long)e->block_id; - if (likely(drbd_bio_uptodate(e->private_bio))) { + if (likely((e->flags & EE_WAS_ERROR) == 0)) { /* quick hack to try to avoid a race against reconfiguration. * a real fix would be much more involved, * introducing more locking mechanisms */ @@ -944,7 +974,7 @@ int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) digest = kmalloc(digest_size, GFP_NOIO); } if (digest) { - drbd_csum(mdev, mdev->csums_tfm, e->private_bio, digest); + drbd_csum_ee(mdev, mdev->csums_tfm, e, digest); eq = !memcmp(digest, di->digest, digest_size); kfree(digest); } @@ -986,14 +1016,14 @@ int w_e_end_ov_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) if (unlikely(cancel)) goto out; - if (unlikely(!drbd_bio_uptodate(e->private_bio))) + if (unlikely((e->flags & EE_WAS_ERROR) != 0)) goto out; digest_size = crypto_hash_digestsize(mdev->verify_tfm); /* FIXME if this allocation fails, online verify will not terminate! */ digest = kmalloc(digest_size, GFP_NOIO); if (digest) { - drbd_csum(mdev, mdev->verify_tfm, e->private_bio, digest); + drbd_csum_ee(mdev, mdev->verify_tfm, e, digest); inc_rs_pending(mdev); ok = drbd_send_drequest_csum(mdev, e->sector, e->size, digest, digest_size, P_OV_REPLY); @@ -1042,11 +1072,11 @@ int w_e_end_ov_reply(struct drbd_conf *mdev, struct drbd_work *w, int cancel) di = (struct digest_info *)(unsigned long)e->block_id; - if (likely(drbd_bio_uptodate(e->private_bio))) { + if (likely((e->flags & EE_WAS_ERROR) == 0)) { digest_size = crypto_hash_digestsize(mdev->verify_tfm); digest = kmalloc(digest_size, GFP_NOIO); if (digest) { - drbd_csum(mdev, mdev->verify_tfm, e->private_bio, digest); + drbd_csum_ee(mdev, mdev->verify_tfm, e, digest); D_ASSERT(digest_size == di->digest_size); eq = !memcmp(digest, di->digest, digest_size); diff --git a/drivers/block/drbd/drbd_wrappers.h b/drivers/block/drbd/drbd_wrappers.h index f93fa11..defdb50 100644 --- a/drivers/block/drbd/drbd_wrappers.h +++ b/drivers/block/drbd/drbd_wrappers.h @@ -18,23 +18,9 @@ static inline void drbd_set_my_capacity(struct drbd_conf *mdev, #define drbd_bio_uptodate(bio) bio_flagged(bio, BIO_UPTODATE) -static inline int drbd_bio_has_active_page(struct bio *bio) -{ - struct bio_vec *bvec; - int i; - - __bio_for_each_segment(bvec, bio, i, 0) { - if (page_count(bvec->bv_page) > 1) - return 1; - } - - return 0; -} - /* bi_end_io handlers */ extern void drbd_md_io_complete(struct bio *bio, int error); -extern void drbd_endio_read_sec(struct bio *bio, int error); -extern void drbd_endio_write_sec(struct bio *bio, int error); +extern void drbd_endio_sec(struct bio *bio, int error); extern void drbd_endio_pri(struct bio *bio, int error); /* -- cgit v1.1 From bb3d000cb99aa0924b78c1ae5f5943484527868a Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Fri, 14 May 2010 19:08:55 +0200 Subject: drbd: allow resync requests to be larger than max_segment_size this should allow for better background resync performance. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_worker.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index d771b1e..91085c1 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -462,7 +462,7 @@ int w_make_resync_request(struct drbd_conf *mdev, unsigned long bit; sector_t sector; const sector_t capacity = drbd_get_capacity(mdev->this_bdev); - int max_segment_size = queue_max_segment_size(mdev->rq_queue); + int max_segment_size; int number, i, size, pe, mx; int align, queued, sndbuf; @@ -488,6 +488,11 @@ int w_make_resync_request(struct drbd_conf *mdev, return 1; } + /* starting with drbd 8.3.8, we can handle multi-bio EEs, + * if it should be necessary */ + max_segment_size = mdev->agreed_pro_version < 94 ? + queue_max_segment_size(mdev->rq_queue) : DRBD_MAX_SEGMENT_SIZE; + mdev->c_sync_rate = calc_resync_rate(mdev); number = SLEEP_TIME * mdev->c_sync_rate / ((BM_BLOCK_SIZE / 1024) * HZ); pe = atomic_read(&mdev->rs_pending_cnt); @@ -552,12 +557,6 @@ next_sector: * * Additionally always align bigger requests, in order to * be prepared for all stripe sizes of software RAIDs. - * - * we _do_ care about the agreed-upon q->max_segment_size - * here, as splitting up the requests on the other side is more - * difficult. the consequence is, that on lvm and md and other - * "indirect" devices, this is dead code, since - * q->max_segment_size will be PAGE_SIZE. */ align = 1; for (;;) { -- cgit v1.1 From a1c88d0d7aa2ef427f78834c9a3b0a673a19dca6 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Fri, 14 May 2010 19:16:41 +0200 Subject: drbd: always use_bmbv, ignore setting Now that the peer may handle multi-bio EEs, we can ignore the peer's limit, and concentrate on the limits of the local IO stack. This is safe accross drbd protocol versions, as our queue_max_sectors() will be adjusted accordingly. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_nl.c | 3 --- drivers/block/drbd/drbd_receiver.c | 6 +++++- drivers/block/drbd/drbd_req.c | 2 +- 3 files changed, 6 insertions(+), 5 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 28ef76b..f20336b 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -704,9 +704,6 @@ void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_seg_s) __mu struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; int max_segments = mdev->ldev->dc.max_bio_bvecs; - if (b->merge_bvec_fn && !mdev->ldev->dc.use_bmbv) - max_seg_s = PAGE_SIZE; - max_seg_s = min(queue_max_sectors(b) * queue_logical_block_size(b), max_seg_s); blk_queue_max_hw_sectors(q, max_seg_s >> 9); diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 388a3e8..a04ec01 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -3011,7 +3011,11 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) ldsc = 1; } - max_seg_s = be32_to_cpu(p->max_segment_size); + if (mdev->agreed_pro_version < 94) + max_seg_s = be32_to_cpu(p->max_segment_size); + else /* drbd 8.3.8 onwards */ + max_seg_s = DRBD_MAX_SEGMENT_SIZE; + if (max_seg_s != queue_max_segment_size(mdev->rq_queue)) drbd_setup_queue_param(mdev, max_seg_s); diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c index d8d9bbf..343e0e6 100644 --- a/drivers/block/drbd/drbd_req.c +++ b/drivers/block/drbd/drbd_req.c @@ -1110,7 +1110,7 @@ int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct } else if (limit && get_ldev(mdev)) { struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; - if (b->merge_bvec_fn && mdev->ldev->dc.use_bmbv) { + if (b->merge_bvec_fn) { backing_limit = b->merge_bvec_fn(b, bvm, bvec); limit = min(limit, backing_limit); } -- cgit v1.1 From 9a25a04c8079725c1b1ab756694a8e0757844b40 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Mon, 10 May 2010 16:42:23 +0200 Subject: drbd: If we detect late that IO got frozen, retry after we thawed. If we detect late (= after grabing mdev->req_lock) that IO got frozen, we return 1 to generic_make_request(), which simply will retry to make a request for that bio. In the subsequent call of generic_make_request() into drbd_make_request_26() we sleep in inc_ap_bio(). Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 4 ++-- drivers/block/drbd/drbd_req.c | 33 ++++++++++++++++++++++++++------- 2 files changed, 28 insertions(+), 9 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 4b97f30..c194348 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -2223,7 +2223,7 @@ static inline int __inc_ap_bio_cond(struct drbd_conf *mdev) /* I'd like to use wait_event_lock_irq, * but I'm not sure when it got introduced, * and not sure when it has 3 or 4 arguments */ -static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two) +static inline void inc_ap_bio(struct drbd_conf *mdev, int count) { /* compare with after_state_ch, * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */ @@ -2245,7 +2245,7 @@ static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two) finish_wait(&mdev->misc_wait, &wait); spin_lock_irq(&mdev->req_lock); } - atomic_add(one_or_two, &mdev->ap_bio_cnt); + atomic_add(count, &mdev->ap_bio_cnt); spin_unlock_irq(&mdev->req_lock); } diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c index 343e0e6..3397f11 100644 --- a/drivers/block/drbd/drbd_req.c +++ b/drivers/block/drbd/drbd_req.c @@ -722,6 +722,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio) struct drbd_request *req; int local, remote; int err = -EIO; + int ret = 0; /* allocate outside of all locks; */ req = drbd_req_new(mdev, bio); @@ -784,7 +785,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio) (mdev->state.pdsk == D_INCONSISTENT && mdev->state.conn >= C_CONNECTED)); - if (!(local || remote)) { + if (!(local || remote) && !mdev->state.susp) { dev_err(DEV, "IO ERROR: neither local nor remote disk\n"); goto fail_free_complete; } @@ -810,6 +811,16 @@ allocate_barrier: /* GOOD, everything prepared, grab the spin_lock */ spin_lock_irq(&mdev->req_lock); + if (mdev->state.susp) { + /* If we got suspended, use the retry mechanism of + generic_make_request() to restart processing of this + bio. In the next call to drbd_make_request_26 + we sleep in inc_ap_bio() */ + ret = 1; + spin_unlock_irq(&mdev->req_lock); + goto fail_free_complete; + } + if (remote) { remote = (mdev->state.pdsk == D_UP_TO_DATE || (mdev->state.pdsk == D_INCONSISTENT && @@ -947,12 +958,14 @@ fail_and_free_req: req->private_bio = NULL; put_ldev(mdev); } - bio_endio(bio, err); + if (!ret) + bio_endio(bio, err); + drbd_req_free(req); dec_ap_bio(mdev); kfree(b); - return 0; + return ret; } /* helper function for drbd_make_request @@ -1065,15 +1078,21 @@ int drbd_make_request_26(struct request_queue *q, struct bio *bio) /* we need to get a "reference count" (ap_bio_cnt) * to avoid races with the disconnect/reconnect/suspend code. - * In case we need to split the bio here, we need to get two references + * In case we need to split the bio here, we need to get three references * atomically, otherwise we might deadlock when trying to submit the * second one! */ - inc_ap_bio(mdev, 2); + inc_ap_bio(mdev, 3); D_ASSERT(e_enr == s_enr + 1); - drbd_make_request_common(mdev, &bp->bio1); - drbd_make_request_common(mdev, &bp->bio2); + while (drbd_make_request_common(mdev, &bp->bio1)) + inc_ap_bio(mdev, 1); + + while (drbd_make_request_common(mdev, &bp->bio2)) + inc_ap_bio(mdev, 1); + + dec_ap_bio(mdev); + bio_pair_release(bp); } return 0; -- cgit v1.1 From 0c3f34516e8c5a1a0ba3585a7777d32bbbdf4ecb Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Mon, 17 May 2010 16:10:43 +0200 Subject: drbd: Create new current UUID as late as possible The choice was to either delay creation of the new UUID until IO got thawed or to delay it until the first IO request. Both are correct, the later is more friendly to users of dual-primary setups, that actually only write on one side. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 9 ++++++++- drivers/block/drbd/drbd_main.c | 25 ++++++++++++++++++++----- drivers/block/drbd/drbd_receiver.c | 11 +++++++++++ 3 files changed, 39 insertions(+), 6 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index c194348..e9654c8 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -943,7 +943,8 @@ struct drbd_conf { struct drbd_work resync_work, unplug_work, md_sync_work, - delay_probe_work; + delay_probe_work, + uuid_work; struct timer_list resync_timer; struct timer_list md_sync_timer; struct timer_list delay_probe_timer; @@ -1068,6 +1069,7 @@ struct drbd_conf { struct timeval dps_time; /* delay-probes-start-time */ unsigned int dp_volume_last; /* send_cnt of last delay probe */ int c_sync_rate; /* current resync rate after delay_probe magic */ + atomic_t new_c_uuid; }; static inline struct drbd_conf *minor_to_mdev(unsigned int minor) @@ -2217,6 +2219,8 @@ static inline int __inc_ap_bio_cond(struct drbd_conf *mdev) return 0; if (test_bit(BITMAP_IO, &mdev->flags)) return 0; + if (atomic_read(&mdev->new_c_uuid)) + return 0; return 1; } @@ -2237,6 +2241,9 @@ static inline void inc_ap_bio(struct drbd_conf *mdev, int count) * to avoid races with the reconnect code, * we need to atomic_inc within the spinlock. */ + if (atomic_read(&mdev->new_c_uuid) && atomic_add_unless(&mdev->new_c_uuid, -1, 1)) + drbd_queue_work_front(&mdev->data.work, &mdev->uuid_work); + spin_lock_irq(&mdev->req_lock); while (!__inc_ap_bio_cond(mdev)) { prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE); diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index d0fabac..c144509 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1217,17 +1217,16 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, mdev->p_uuid = NULL; if (get_ldev(mdev)) { if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && - mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { - drbd_uuid_new_current(mdev); - drbd_send_uuids(mdev); - } + mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) + atomic_set(&mdev->new_c_uuid, 2); put_ldev(mdev); } } if (ns.pdsk < D_INCONSISTENT && get_ldev(mdev)) { + /* Diskless peer becomes primary or got connected do diskless, primary peer. */ if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0) - drbd_uuid_new_current(mdev); + atomic_set(&mdev->new_c_uuid, 2); /* D_DISKLESS Peer becomes secondary */ if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) @@ -1351,6 +1350,19 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, drbd_md_sync(mdev); } +static int w_new_current_uuid(struct drbd_conf *mdev, struct drbd_work *w, int cancel) +{ + if (get_ldev(mdev)) { + drbd_uuid_new_current(mdev); + drbd_send_uuids(mdev); + drbd_md_sync(mdev); + put_ldev(mdev); + } + atomic_dec(&mdev->new_c_uuid); + wake_up(&mdev->misc_wait); + + return 1; +} static int drbd_thread_setup(void *arg) { @@ -2691,6 +2703,7 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) atomic_set(&mdev->net_cnt, 0); atomic_set(&mdev->packet_seq, 0); atomic_set(&mdev->pp_in_use, 0); + atomic_set(&mdev->new_c_uuid, 0); mutex_init(&mdev->md_io_mutex); mutex_init(&mdev->data.mutex); @@ -2721,12 +2734,14 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) INIT_LIST_HEAD(&mdev->bm_io_work.w.list); INIT_LIST_HEAD(&mdev->delay_probes); INIT_LIST_HEAD(&mdev->delay_probe_work.list); + INIT_LIST_HEAD(&mdev->uuid_work.list); mdev->resync_work.cb = w_resync_inactive; mdev->unplug_work.cb = w_send_write_hint; mdev->md_sync_work.cb = w_md_sync; mdev->bm_io_work.w.cb = w_bitmap_io; mdev->delay_probe_work.cb = w_delay_probes; + mdev->uuid_work.cb = w_new_current_uuid; init_timer(&mdev->resync_timer); init_timer(&mdev->md_sync_timer); init_timer(&mdev->delay_probe_timer); diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index a04ec01..461d987 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -1150,6 +1150,17 @@ int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, unsigned n_bios = 0; unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT; + if (atomic_read(&mdev->new_c_uuid)) { + if (atomic_add_unless(&mdev->new_c_uuid, -1, 1)) { + drbd_uuid_new_current(mdev); + drbd_md_sync(mdev); + + atomic_dec(&mdev->new_c_uuid); + wake_up(&mdev->misc_wait); + } + wait_event(mdev->misc_wait, !atomic_read(&mdev->new_c_uuid)); + } + /* In most cases, we will only need one bio. But in case the lower * level restrictions happen to be different at this offset on this * side than those of the sending peer, we may need to submit the -- cgit v1.1 From 2db4e42eaceabec42f738f3895300632cd375e67 Mon Sep 17 00:00:00 2001 From: Julia Lawall Date: Thu, 13 May 2010 22:02:21 +0200 Subject: drivers/block/drbd: Use kzalloc Use kzalloc rather than the combination of kmalloc and memset. The semantic patch that makes this change is as follows: (http://coccinelle.lip6.fr/) // @@ expression x,size,flags; statement S; @@ -x = kmalloc(size,flags); +x = kzalloc(size,flags); if (x == NULL) S -memset(x, 0, size); // Signed-off-by: Julia Lawall Signed-off-by: Philipp Reisner Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_nl.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index f20336b..632e324 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -1196,13 +1196,12 @@ static int drbd_nl_net_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, } /* allocation not in the IO path, cqueue thread context */ - new_conf = kmalloc(sizeof(struct net_conf), GFP_KERNEL); + new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); if (!new_conf) { retcode = ERR_NOMEM; goto fail; } - memset(new_conf, 0, sizeof(struct net_conf)); new_conf->timeout = DRBD_TIMEOUT_DEF; new_conf->try_connect_int = DRBD_CONNECT_INT_DEF; new_conf->ping_int = DRBD_PING_INT_DEF; -- cgit v1.1 From 4cb2ea28c55cf5e5ef83aec535099ffce3c583df Mon Sep 17 00:00:00 2001 From: john cooper Date: Thu, 25 Mar 2010 01:33:33 -0400 Subject: Add virtio disk identification support Add virtio-blk device id (s/n) support via virtio request. Signed-off-by: john cooper Signed-off-by: Rusty Russell --- drivers/block/virtio_blk.c | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+) (limited to 'drivers/block') diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index 2138a7a..759dee8 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -70,6 +70,8 @@ static void blk_done(struct virtqueue *vq) vbr->req->sense_len = vbr->in_hdr.sense_len; vbr->req->errors = vbr->in_hdr.errors; } + if (blk_special_request(vbr->req)) + vbr->req->errors = (error != 0); __blk_end_request_all(vbr->req, error); list_del(&vbr->list); @@ -103,6 +105,11 @@ static bool do_req(struct request_queue *q, struct virtio_blk *vblk, vbr->out_hdr.sector = 0; vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); break; + case REQ_TYPE_SPECIAL: + vbr->out_hdr.type = VIRTIO_BLK_T_GET_ID; + vbr->out_hdr.sector = 0; + vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); + break; case REQ_TYPE_LINUX_BLOCK: if (req->cmd[0] == REQ_LB_OP_FLUSH) { vbr->out_hdr.type = VIRTIO_BLK_T_FLUSH; @@ -189,6 +196,29 @@ static void virtblk_prepare_flush(struct request_queue *q, struct request *req) req->cmd[0] = REQ_LB_OP_FLUSH; } +/* return id (s/n) string for *disk to *id_str + */ +static int virtblk_get_id(struct gendisk *disk, char *id_str) +{ + struct virtio_blk *vblk = disk->private_data; + struct request *req; + struct bio *bio; + + bio = bio_map_kern(vblk->disk->queue, id_str, VIRTIO_BLK_ID_BYTES, + GFP_KERNEL); + if (IS_ERR(bio)) + return PTR_ERR(bio); + + req = blk_make_request(vblk->disk->queue, bio, GFP_KERNEL); + if (IS_ERR(req)) { + bio_put(bio); + return PTR_ERR(req); + } + + req->cmd_type = REQ_TYPE_SPECIAL; + return blk_execute_rq(vblk->disk->queue, vblk->disk, req, false); +} + static int virtblk_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, unsigned long data) { -- cgit v1.1 From 234f2725a5d03f78539f1d36cb32f2c4f9b1822c Mon Sep 17 00:00:00 2001 From: john cooper Date: Thu, 25 Mar 2010 01:34:02 -0400 Subject: Add virtio disk identification ioctl Return serial string to the guest application via ioctl driver call. Note this form of interface to the guest userland was the consensus when the prior version using the ATA_IDENTIFY came under dispute. Signed-off-by: john cooper Signed-off-by: Rusty Russell --- drivers/block/virtio_blk.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'drivers/block') diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index 759dee8..67dcb19 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -225,6 +225,16 @@ static int virtblk_ioctl(struct block_device *bdev, fmode_t mode, struct gendisk *disk = bdev->bd_disk; struct virtio_blk *vblk = disk->private_data; + if (cmd == 'VBID') { + void __user *usr_data = (void __user *)data; + char id_str[VIRTIO_BLK_ID_BYTES]; + int err; + + err = virtblk_get_id(disk, id_str); + if (!err && copy_to_user(usr_data, id_str, VIRTIO_BLK_ID_BYTES)) + err = -EFAULT; + return err; + } /* * Only allow the generic SCSI ioctls if the host can support it. */ -- cgit v1.1 From bdb4a1305723f985249210a803105dbc48e86b64 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Wed, 19 May 2010 22:15:40 -0600 Subject: virtio_blk: remove multichar constant. drivers/block/virtio_blk.c:228:13: warning: multi-character character constant Signed-off-by: Rusty Russell Cc: john cooper --- drivers/block/virtio_blk.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index 67dcb19..e32b24b 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -225,7 +225,7 @@ static int virtblk_ioctl(struct block_device *bdev, fmode_t mode, struct gendisk *disk = bdev->bd_disk; struct virtio_blk *vblk = disk->private_data; - if (cmd == 'VBID') { + if (cmd == 0x56424944) { /* 'VBID' */ void __user *usr_data = (void __user *)data; char id_str[VIRTIO_BLK_ID_BYTES]; int err; -- cgit v1.1 From 09ec6b69d2b97d6fca16cfe91b4634506f4db0a7 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Mon, 12 Apr 2010 16:18:36 +0300 Subject: virtio_blk: use virtqueue_xxx wrappers Switch virtio_blk to new virtqueue_xxx wrappers. Signed-off-by: Michael S. Tsirkin Signed-off-by: Rusty Russell --- drivers/block/virtio_blk.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index e32b24b..83fa09a 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -50,7 +50,7 @@ static void blk_done(struct virtqueue *vq) unsigned long flags; spin_lock_irqsave(&vblk->lock, flags); - while ((vbr = vblk->vq->vq_ops->get_buf(vblk->vq, &len)) != NULL) { + while ((vbr = virtqueue_get_buf(vblk->vq, &len)) != NULL) { int error; switch (vbr->status) { @@ -158,7 +158,7 @@ static bool do_req(struct request_queue *q, struct virtio_blk *vblk, } } - if (vblk->vq->vq_ops->add_buf(vblk->vq, vblk->sg, out, in, vbr) < 0) { + if (virtqueue_add_buf(vblk->vq, vblk->sg, out, in, vbr) < 0) { mempool_free(vbr, vblk->pool); return false; } @@ -187,7 +187,7 @@ static void do_virtblk_request(struct request_queue *q) } if (issued) - vblk->vq->vq_ops->kick(vblk->vq); + virtqueue_kick(vblk->vq); } static void virtblk_prepare_flush(struct request_queue *q, struct request *req) -- cgit v1.1 From c3473c63542d53740f175f3a515257ae159e998b Mon Sep 17 00:00:00 2001 From: David Zeuthen Date: Mon, 3 May 2010 14:08:59 +0200 Subject: generate "change" uevent for loop device Recent udev versions probe loop devices for filesystems meaning that the /dev/disk hierarchy may contain useful entries such as $ ls -l /dev/disk/by-label/Fedora-12-x86_64-Live lrwxrwxrwx 1 root root 11 Mar 11 13:41 /dev/disk/by-label/Fedora-12-x86_64-Live -> ../../loop0 Unfortunately, no "change" uevent is generated when the loop device is detached so the symlink persists. Additionally, no "change" uevent is guaranteed to be generated when attaching an fd or changing capacity. For example, user space could open the loop device O_RDONLY (in fact, recent util-linux-ng does this) so udev's OPTIONS+="watch" machinery may not trigger the "change" uevent. This patch ensures that the "change" uevent is generated in all of these cases. As a result, the /dev/disk hierarchy works as expected for loop devices. Signed-off-by: David Zeuthen Acked-by: Kay Sievers Signed-off-by: Greg Kroah-Hartman --- drivers/block/loop.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 8546d12..a90e83c 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -835,6 +835,8 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode, set_capacity(lo->lo_disk, size); bd_set_size(bdev, size << 9); + /* let user-space know about the new size */ + kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); set_blocksize(bdev, lo_blocksize); @@ -858,6 +860,7 @@ out_clr: set_capacity(lo->lo_disk, 0); invalidate_bdev(bdev); bd_set_size(bdev, 0); + kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); mapping_set_gfp_mask(mapping, lo->old_gfp_mask); lo->lo_state = Lo_unbound; out_putf: @@ -944,8 +947,11 @@ static int loop_clr_fd(struct loop_device *lo, struct block_device *bdev) if (bdev) invalidate_bdev(bdev); set_capacity(lo->lo_disk, 0); - if (bdev) + if (bdev) { bd_set_size(bdev, 0); + /* let user-space know about this change */ + kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); + } mapping_set_gfp_mask(filp->f_mapping, gfp); lo->lo_state = Lo_unbound; /* This is safe: open() is still holding a reference. */ @@ -1189,6 +1195,8 @@ static int loop_set_capacity(struct loop_device *lo, struct block_device *bdev) sz <<= 9; mutex_lock(&bdev->bd_mutex); bd_set_size(bdev, sz); + /* let user-space know about the new size */ + kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); mutex_unlock(&bdev->bd_mutex); out: -- cgit v1.1 From 31a31dccdd308d5ec7f9a1197a7875a246a348dc Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 19 May 2010 16:28:28 +0200 Subject: drbd: Do not Oops when C_STANDALONE when uuid gets generated Got introduces with commit 0c3f34516e8c5a1a0ba3585a7777d32bbbdf4ecb Author: Philipp Reisner Date: Mon May 17 16:10:43 2010 +0200 drbd: Create new current UUID as late as possible Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_main.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index c144509..7e057b0 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1354,7 +1354,10 @@ static int w_new_current_uuid(struct drbd_conf *mdev, struct drbd_work *w, int c { if (get_ldev(mdev)) { drbd_uuid_new_current(mdev); - drbd_send_uuids(mdev); + if (get_net_conf(mdev)) { + drbd_send_uuids(mdev); + put_net_conf(mdev); + } drbd_md_sync(mdev); put_ldev(mdev); } -- cgit v1.1 From 4604d6366859f781ad16c07a6c65b16fc96e26c5 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 19 May 2010 17:37:02 +0200 Subject: drbd: Ensure to not trigger late-new-UUID creation multiple times Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_main.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 7e057b0..a949fc4 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1217,7 +1217,8 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, mdev->p_uuid = NULL; if (get_ldev(mdev)) { if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && - mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) + mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE && + !atomic_read(&mdev->new_c_uuid)) atomic_set(&mdev->new_c_uuid, 2); put_ldev(mdev); } @@ -1225,7 +1226,8 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, if (ns.pdsk < D_INCONSISTENT && get_ldev(mdev)) { /* Diskless peer becomes primary or got connected do diskless, primary peer. */ - if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0) + if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0 && + !atomic_read(&mdev->new_c_uuid)) atomic_set(&mdev->new_c_uuid, 2); /* D_DISKLESS Peer becomes secondary */ @@ -1353,12 +1355,14 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, static int w_new_current_uuid(struct drbd_conf *mdev, struct drbd_work *w, int cancel) { if (get_ldev(mdev)) { - drbd_uuid_new_current(mdev); - if (get_net_conf(mdev)) { - drbd_send_uuids(mdev); - put_net_conf(mdev); + if (mdev->ldev->md.uuid[UI_BITMAP] == 0) { + drbd_uuid_new_current(mdev); + if (get_net_conf(mdev)) { + drbd_send_uuids(mdev); + put_net_conf(mdev); + } + drbd_md_sync(mdev); } - drbd_md_sync(mdev); put_ldev(mdev); } atomic_dec(&mdev->new_c_uuid); -- cgit v1.1 From fc8ce1941d668c70e57a07f13f5a63e73e5dbff3 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 20 May 2010 10:04:17 +0200 Subject: drbd: Fix: Do not detach, if a bio with a barrier fails Introduced a few days ago: commit 45bb912bd5ea4d2b3a270a93cbdf767a0e2df6f5 Author: Lars Ellenberg Date: Fri May 14 17:10:48 2010 +0200 Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_worker.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index 91085c1..15c9620 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -127,7 +127,7 @@ static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(lo drbd_bump_write_ordering(mdev, WO_bdev_flush); spin_lock_irqsave(&mdev->req_lock, flags); list_del(&e->w.list); - e->flags |= EE_RESUBMITTED; + e->flags = (e->flags & ~EE_WAS_ERROR) | EE_RESUBMITTED; e->w.cb = w_e_reissue; /* put_ldev actually happens below, once we come here again. */ __release(local); -- cgit v1.1 From 23ce422748def9652fdc019f740cd7900fa2fe3b Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 20 May 2010 13:35:31 +0200 Subject: drbd: Null pointer deref fix to the large "multi bio rewrite" Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_receiver.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 461d987..bc9ab7f 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -98,6 +98,10 @@ static struct page *page_chain_del(struct page **head, int n) BUG_ON(!head); page = *head; + + if (!page) + return NULL; + while (page) { tmp = page_chain_next(page); if (--n == 0) -- cgit v1.1 From 4e23a59ed1c5f12e14f7899855f8379b3d42e578 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Thu, 20 May 2010 14:45:07 +0200 Subject: drbd: Do not free p_uuid early, this is done in the exit code of the receiver Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_main.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index a949fc4..be2d2da 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1213,8 +1213,6 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, && (ns.pdsk < D_INCONSISTENT || ns.pdsk == D_UNKNOWN || ns.pdsk == D_OUTDATED)) { - kfree(mdev->p_uuid); - mdev->p_uuid = NULL; if (get_ldev(mdev)) { if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE && -- cgit v1.1 From 8018ab057480974e7f26a387bf4ce040e9a5f6f1 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 22 Mar 2010 17:32:25 +0100 Subject: sanitize vfs_fsync calling conventions Now that the last user passing a NULL file pointer is gone we can remove the redundant dentry argument and associated hacks inside vfs_fsynmc_range. The next step will be removig the dentry argument from ->fsync, but given the luck with the last round of method prototype changes I'd rather defer this until after the main merge window. Signed-off-by: Christoph Hellwig Signed-off-by: Al Viro --- drivers/block/loop.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/loop.c b/drivers/block/loop.c index a90e83c..6120922 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -485,7 +485,7 @@ static int do_bio_filebacked(struct loop_device *lo, struct bio *bio) goto out; } - ret = vfs_fsync(file, file->f_path.dentry, 0); + ret = vfs_fsync(file, 0); if (unlikely(ret)) { ret = -EIO; goto out; @@ -495,7 +495,7 @@ static int do_bio_filebacked(struct loop_device *lo, struct bio *bio) ret = lo_send(lo, bio, pos); if (barrier && !ret) { - ret = vfs_fsync(file, file->f_path.dentry, 0); + ret = vfs_fsync(file, 0); if (unlikely(ret)) ret = -EIO; } -- cgit v1.1