[PATCH 4/4] btrfs: add discard_compat support

[PATCH 4/4] btrfs: add discard_compat support

[Dmitry Monakhov]

If any device in the list has no native discard support we add
discard compat support. Devices with native discard support still
getting real discard requests.

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
---
 fs/btrfs/ctree.h       |    1 +
 fs/btrfs/disk-io.c     |    8 ++++++++
 fs/btrfs/extent-tree.c |   14 ++++++++------
 fs/btrfs/super.c       |    8 +++++++-
 fs/btrfs/volumes.c     |    6 ++++++
 fs/btrfs/volumes.h     |    6 +++++-
 6 files changed, 35 insertions(+), 8 deletions(-)

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9f806dd..54854d1 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1161,6 +1161,7 @@ struct btrfs_root {
 #define BTRFS_MOUNT_SSD_SPREAD		(1 << 8)
 #define BTRFS_MOUNT_NOSSD		(1 << 9)
 #define BTRFS_MOUNT_DISCARD		(1 << 10)
+#define BTRFS_MOUNT_DISCARD_COMPAT	(1 << 11)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 009e3bd..c7d4812 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1945,6 +1945,14 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		       "mode\n");
 		btrfs_set_opt(fs_info->mount_opt, SSD);
 	}
+	if (btrfs_test_opt(tree_root, DISCARD) &&
+	    fs_info->fs_devices->discard_compat) {
+		printk(KERN_INFO "Btrfs detected devices without native discard"
+			" support, enabling discard_compat mode mode\n");
+		btrfs_clear_opt(fs_info->mount_opt, DISCARD);
+		btrfs_set_opt(fs_info->mount_opt, DISCARD_COMPAT);
+	}
+
 
 	if (btrfs_super_log_root(disk_super) != 0) {
 		u64 bytenr = btrfs_super_log_root(disk_super);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 432a2da..b4c3124 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1585,20 +1585,21 @@ static int remove_extent_backref(struct btrfs_trans_handle *trans,
 }
 
 static void btrfs_issue_discard(struct block_device *bdev,
-				u64 start, u64 len)
+				u64 start, u64 len, int do_compat)
 {
 	blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL,
-			     DISCARD_FL_BARRIER);
+			     DISCARD_FL_BARRIER |
+				(do_compat ? DISCARD_FL_COMPAT : 0));
 }
 
 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
 				u64 num_bytes)
 {
-	int ret;
+	int ret, do_compat = btrfs_test_opt(root, DISCARD_COMPAT);
 	u64 map_length = num_bytes;
 	struct btrfs_multi_bio *multi = NULL;
 
-	if (!btrfs_test_opt(root, DISCARD))
+	if (!btrfs_test_opt(root, DISCARD) && !do_compat)
 		return 0;
 
 	/* Tell the block device(s) that the sectors can be discarded */
@@ -1614,7 +1615,8 @@ static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
 		for (i = 0; i < multi->num_stripes; i++, stripe++) {
 			btrfs_issue_discard(stripe->dev->bdev,
 					    stripe->physical,
-					    map_length);
+					    map_length,
+					    do_compat);
 		}
 		kfree(multi);
 	}
@@ -3700,7 +3702,7 @@ static int pin_down_bytes(struct btrfs_trans_handle *trans,
 	 * individual btree blocks isn't a good plan.  Just
 	 * pin everything in discard mode.
 	 */
-	if (btrfs_test_opt(root, DISCARD))
+	if (btrfs_test_opt(root, DISCARD) || btrfs_test_opt(root, DISCARD_COMPAT))
 		goto pinit;
 
 	buf = btrfs_find_tree_block(root, bytenr, num_bytes);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 3f9b457..fe7ccf4 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -67,7 +67,7 @@ enum {
 	Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
 	Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl,
 	Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
-	Opt_discard, Opt_err,
+	Opt_discard, Opt_discard_compat, Opt_err,
 };
 
 static match_table_t tokens = {
@@ -90,6 +90,7 @@ static match_table_t tokens = {
 	{Opt_flushoncommit, "flushoncommit"},
 	{Opt_ratio, "metadata_ratio=%d"},
 	{Opt_discard, "discard"},
+	{Opt_discard_compat, "discard_compat"},
 	{Opt_err, NULL},
 };
 
@@ -263,6 +264,9 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 		case Opt_discard:
 			btrfs_set_opt(info->mount_opt, DISCARD);
 			break;
+		case Opt_discard_compat:
+			btrfs_set_opt(info->mount_opt, DISCARD_COMPAT);
+			break;
 		case Opt_err:
 			printk(KERN_INFO "btrfs: unrecognized mount option "
 			       "'%s'\n", p);
@@ -459,6 +463,8 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_puts(seq, ",flushoncommit");
 	if (btrfs_test_opt(root, DISCARD))
 		seq_puts(seq, ",discard");
+	if (btrfs_test_opt(root, DISCARD_COMPAT))
+		seq_puts(seq, ",discard_compat");
 	if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 		seq_puts(seq, ",noacl");
 	return 0;
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 220dad5..02e18c8 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -621,6 +621,9 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		if (!blk_queue_nonrot(bdev_get_queue(bdev)))
 			fs_devices->rotating = 1;
 
+		if (!blk_queue_discard(bdev_get_queue(bdev)))
+			fs_devices->discard_compat = 1;
+
 		fs_devices->open_devices++;
 		if (device->writeable) {
 			fs_devices->rw_devices++;
@@ -1522,6 +1525,9 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	if (!blk_queue_nonrot(bdev_get_queue(bdev)))
 		root->fs_info->fs_devices->rotating = 1;
 
+	if (!blk_queue_discard(bdev_get_queue(bdev)))
+		root->fs_info->fs_devices->discard_compat = 1;
+
 	total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
 	btrfs_set_super_total_bytes(&root->fs_info->super_copy,
 				    total_bytes + device->total_bytes);
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 31b0fab..ceb66f0 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -116,7 +116,11 @@ struct btrfs_fs_devices {
 	/* set when we find or add a device that doesn't have the
 	 * nonrot flag set
 	 */
-	int rotating;
+	unsigned int rotating:1;
+	/* set then we find or add a device that doesn't have the
+	 * DISCARD flags set
+	 */
+	unsigned int discard_compat:1;
 };
 
 struct btrfs_bio_stripe {
-- 
1.6.6

Re: [PATCH 4/4] btrfs: add discard_compat support

[Dmitry Monakhov]

Dmitry Monakhov <dmonakhov@openvz.org> writes:

> If any device in the list has no native discard support we add
> discard compat support. Devices with native discard support still
> getting real discard requests.
Note: Seems what enabling discard requests result in triggering
hidden bug. Orphan list corruption.

  ------------[ cut here ]------------
 WARNING: at lib/list_debug.c:26 __list_add+0x3f/0x81()
 Hardware name:         
 list_add corruption. next->prev should be prev (ffff88012029eb20), but was ffff88012239c730. (next=ffff8801223bc348).
 Modules linked in: btrfs zlib_deflate libcrc32c cpufreq_ondemand acpi_cpufreq freq_table ppdev parport_pc parport pcspkr [last unloaded: btrfs]
 Pid: 1254, comm: fsstress Not tainted 2.6.33-rc5 #3
 Call Trace:
 [<ffffffff81038f79>] warn_slowpath_common+0x7c/0x94
 [<ffffffff81038fe8>] warn_slowpath_fmt+0x41/0x43
 [<ffffffffa00df87a>] ? btrfs_unlink_inode+0x243/0x258 [btrfs]
 [<ffffffff811c0e22>] __list_add+0x3f/0x81
 [<ffffffffa00dfac7>] btrfs_orphan_add+0x61/0x80 [btrfs]
 [<ffffffffa00e015a>] btrfs_unlink+0xb2/0xf2 [btrfs]
 [<ffffffff810ec5a3>] vfs_unlink+0x67/0xa2
 [<ffffffff810eafec>] ? lookup_hash+0x36/0x3a
 [<ffffffff810edd97>] do_unlinkat+0xcd/0x15b
 [<ffffffff810eabc4>] ? path_put+0x22/0x27
 [<ffffffff8107edac>] ? audit_syscall_entry+0x11e/0x14a
 [<ffffffff810ede3b>] sys_unlink+0x16/0x18
 [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b
 ---[ end trace 18d84c057a649cf7 ]---
 ------------[ cut here ]------------
 kernel BUG at fs/btrfs/inode.c:3320!
 invalid opcode: 0000 [#1] SMP 
 last sysfs file: /sys/devices/pci0000:00/0000:00:1f.2/host0/target0:0:0/0:0:0:0/block/sda/uevent
 CPU 3 
 Pid: 1250, comm: fsstress Tainted: G        W  2.6.33-rc5 #3 DH55TC                /        
 RIP: 0010:[<ffffffffa00e071b>]  [<ffffffffa00e071b>] btrfs_setattr+0x101/0x244 [btrfs]
 RSP: 0018:ffff88012095de08  EFLAGS: 00010292
 RAX: 0000000000000021 RBX: ffff8801223b4bd8 RCX: 000000000000711a
 RDX: 0000000000000000 RSI: 0000000000000046 RDI: ffffffff81730984
 RBP: ffff88012095de48 R08: ffff88012095ddb4 R09: 0000000000000000
 R10: 0000000000000001 R11: ffff88012095dd00 R12: ffff88012095ded8
 R13: ffff88012029e800 R14: ffff88012233d140 R15: 0000000000000000
 FS:  00007f313c3c6700(0000) GS:ffff8800282c0000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007fffa0ac3f4c CR3: 00000001212c7000 CR4: 00000000000006e0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
 Process fsstress (pid: 1250, threadinfo ffff88012095c000, task ffff880122c68000)
 Stack:
 ffff8801223b4cb0 ffff88011c452180 ffff88012095de28 ffff88012095ded8
 <0> ffff88011c452180 ffff8801223b4bd8 0000000000000008 0000000000000008
 <0> ffff88012095deb8 ffffffff810f5e52 ffff88012095df38 ffffffff810ee418
 Call Trace:
 [<ffffffff810f5e52>] notify_change+0x17b/0x2c9
 [<ffffffff810ee418>] ? user_path_at+0x64/0x93
 [<ffffffff810e1e0d>] do_truncate+0x6a/0x87
 [<ffffffff810ea962>] ? generic_permission+0x1c/0x9f
 [<ffffffff810eaa5e>] ? get_write_access+0x1d/0x48
 [<ffffffff810e1f46>] sys_truncate+0x11c/0x160
 [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b
 Code: e0 48 89 de 4c 89 f7 49 89 46 20 e8 66 f3 ff ff 85 c0 74 1b 89 c2 48 c7 c6 e0 b4 10 a0 48 c7 c7 9c d1 10 a0 31 c0 e8 7c 5c 25 e1 <0f> 0b eb 3d 49 8b 46 10 4c 89 ee 4c 89 f7 48 89 45 c8 e8 47 7d 
 RIP  [<ffffffffa00e071b>] btrfs_setattr+0x101/0x244 [btrfs]
 RSP <ffff88012095de08>
 ---[ end trace 18d84c057a649cf8 ]---