Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Johannes Thumshirn]

Hi Joanne,

After commit aa35dd5cbc06 ("iomap: fix invalid folio access after 
folio_end_read()") my zoned btrfs test setup hangs. I've bisected it to 
this commit and reverting fixes my problem. The last thing I see in 
dmesg is:

[    9.387175] ------------[ cut here ]------------
[    9.387320] WARNING: fs/iomap/buffered-io.c:487 at 
iomap_read_end+0x11c/0x140, CPU#5: (udev-worker)/463
[    9.387431] Modules linked in:
[    9.387502] CPU: 5 UID: 0 PID: 463 Comm: (udev-worker) Not tainted 
6.19.0-rc1+ #385 PREEMPT(full)
[    9.387626] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), 
BIOS 1.17.0-9.fc43 06/10/2025
[    9.387810] RIP: 0010:iomap_read_end+0x11c/0x140
[    9.387886] Code: 00 48 89 ef 48 3b 04 24 0f 94 04 24 44 0f b6 34 24 
e8 b8 88 69 00 48 89 df 48 83 c4 08 41 0f b6 f6 5b 5d 41 5e e9 54 e7 e8 
ff <0f> 0b e9 48 ff ff ff ba 00 10 00 00 eb 9d 0f 0b e9 53 ff ff ff 0f
[    9.388096] RSP: 0018:ffffc90000e279c0 EFLAGS: 00010206
[    9.388178] RAX: ffff888111c6b800 RBX: ffffea0004363900 RCX: 
0000000000000000
[    9.388281] RDX: 0000000000000000 RSI: 0000000000000400 RDI: 
ffffea0004363900
[    9.388386] RBP: 0000000000000000 R08: 0000000000000000 R09: 
0000000000000001
[    9.388491] R10: ffffc90000e279a0 R11: ffff888111c6c168 R12: 
ffffffff81e5cdb0
[    9.388585] R13: ffffc90000e27c58 R14: ffffea0004363900 R15: 
ffffc90000e27c58
[    9.388690] FS:  00007f6e03fdfc00(0000) GS:ffff8882b4c13000(0000) 
knlGS:0000000000000000
[    9.388793] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    9.388873] CR2: 00007f6e03280000 CR3: 00000001106d7006 CR4: 
0000000000770eb0
[    9.388967] PKRU: 55555554
[    9.389006] Call Trace:
[    9.389044]  <TASK>
[    9.389087]  iomap_readahead+0x23c/0x2e0
[    9.389167]  blkdev_readahead+0x3d/0x50
[    9.389222]  read_pages+0x56/0x200
[    9.389277]  ? __folio_batch_add_and_move+0x1cf/0x2d0
[    9.389354]  page_cache_ra_unbounded+0x1db/0x2c0
[    9.389423]  force_page_cache_ra+0x96/0xb0
[    9.389470]  filemap_get_pages+0x12f/0x490
[    9.389532]  filemap_read+0xed/0x400
[    9.389590]  ? lock_acquire+0xd5/0x2b0
[    9.389633]  ? blkdev_read_iter+0x6b/0x180
[    9.389678]  ? lock_acquire+0xe5/0x2b0
[    9.389720]  ? lock_is_held_type+0xcd/0x130
[    9.389761]  ? find_held_lock+0x2b/0x80
[    9.389813]  ? lock_acquired+0x1e9/0x3c0
[    9.389864]  blkdev_read_iter+0x79/0x180
[    9.389911]  ? local_clock_noinstr+0x17/0x110
[    9.389975]  vfs_read+0x240/0x340
[    9.390033]  ksys_read+0x61/0xd0
[    9.390083]  do_syscall_64+0x74/0x3a0
[    9.390143]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
[    9.390204] RIP: 0033:0x7f6e048c5c5e
[    9.390255] Code: 4d 89 d8 e8 34 bd 00 00 4c 8b 5d f8 41 8b 93 08 03 
00 00 59 5e 48 83 f8 fc 74 11 c9 c3 0f 1f 80 00 00 00 00 48 8b 45 10 0f 
05 <c9> c3 83 e2 39 83 fa 08 75 e7 e8 13 ff ff ff 0f 1f 00 f3 0f 1e fa
[    9.390447] RSP: 002b:00007ffdf50cded0 EFLAGS: 00000202 ORIG_RAX: 
0000000000000000
[    9.390529] RAX: ffffffffffffffda RBX: 000000013aa55200 RCX: 
00007f6e048c5c5e
[    9.390619] RDX: 0000000000000200 RSI: 00007f6e0327f000 RDI: 
0000000000000014
[    9.390704] RBP: 00007ffdf50cdee0 R08: 0000000000000000 R09: 
0000000000000000
[    9.390789] R10: 0000000000000000 R11: 0000000000000202 R12: 
0000000000000000
[    9.390866] R13: 000055d5f99fa270 R14: 000055d5f96e5cb0 R15: 
000055d5f96e5cc8
[    9.390967]  </TASK>
[    9.390997] irq event stamp: 54269
[    9.391044] hardirqs last  enabled at (54279): [<ffffffff8138ac02>] 
__up_console_sem+0x52/0x60
[    9.391141] hardirqs last disabled at (54288): [<ffffffff8138abe7>] 
__up_console_sem+0x37/0x60
[    9.391240] softirqs last  enabled at (53796): [<ffffffff81304768>] 
irq_exit_rcu+0x78/0x110
[    9.391327] softirqs last disabled at (53787): [<ffffffff81304768>] 
irq_exit_rcu+0x78/0x110
[    9.391420] ---[ end trace 0000000000000000 ]---

I haven't debugged this further yet, maybe you have an idea what 
could've caused it. On my side it's trivial to reproduce, so if you 
can't reproduce it just yell.

Byte,

     Johannes

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Joanne Koong]

On Mon, Mar 9, 2026 at 9:37 AM Johannes Thumshirn
<Johannes.Thumshirn@wdc.com> wrote:
>
> Hi Joanne,
>
> After commit aa35dd5cbc06 ("iomap: fix invalid folio access after
> folio_end_read()") my zoned btrfs test setup hangs. I've bisected it to
> this commit and reverting fixes my problem. The last thing I see in
> dmesg is:
>
> [    9.387175] ------------[ cut here ]------------
> [    9.387320] WARNING: fs/iomap/buffered-io.c:487 at
> iomap_read_end+0x11c/0x140, CPU#5: (udev-worker)/463
> [    9.387431] Modules linked in:
> [    9.387502] CPU: 5 UID: 0 PID: 463 Comm: (udev-worker) Not tainted
> 6.19.0-rc1+ #385 PREEMPT(full)
> [    9.387626] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
> BIOS 1.17.0-9.fc43 06/10/2025
> [    9.387810] RIP: 0010:iomap_read_end+0x11c/0x140
> [    9.387886] Code: 00 48 89 ef 48 3b 04 24 0f 94 04 24 44 0f b6 34 24
> e8 b8 88 69 00 48 89 df 48 83 c4 08 41 0f b6 f6 5b 5d 41 5e e9 54 e7 e8
> ff <0f> 0b e9 48 ff ff ff ba 00 10 00 00 eb 9d 0f 0b e9 53 ff ff ff 0f
> [    9.388096] RSP: 0018:ffffc90000e279c0 EFLAGS: 00010206
> [    9.388178] RAX: ffff888111c6b800 RBX: ffffea0004363900 RCX:
> 0000000000000000
> [    9.388281] RDX: 0000000000000000 RSI: 0000000000000400 RDI:
> ffffea0004363900
> [    9.388386] RBP: 0000000000000000 R08: 0000000000000000 R09:
> 0000000000000001
> [    9.388491] R10: ffffc90000e279a0 R11: ffff888111c6c168 R12:
> ffffffff81e5cdb0
> [    9.388585] R13: ffffc90000e27c58 R14: ffffea0004363900 R15:
> ffffc90000e27c58
> [    9.388690] FS:  00007f6e03fdfc00(0000) GS:ffff8882b4c13000(0000)
> knlGS:0000000000000000
> [    9.388793] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [    9.388873] CR2: 00007f6e03280000 CR3: 00000001106d7006 CR4:
> 0000000000770eb0
> [    9.388967] PKRU: 55555554
> [    9.389006] Call Trace:
> [    9.389044]  <TASK>
> [    9.389087]  iomap_readahead+0x23c/0x2e0
> [    9.389167]  blkdev_readahead+0x3d/0x50
> [    9.389222]  read_pages+0x56/0x200
> [    9.389277]  ? __folio_batch_add_and_move+0x1cf/0x2d0
> [    9.389354]  page_cache_ra_unbounded+0x1db/0x2c0
> [    9.389423]  force_page_cache_ra+0x96/0xb0
> [    9.389470]  filemap_get_pages+0x12f/0x490
> [    9.389532]  filemap_read+0xed/0x400
> [    9.389590]  ? lock_acquire+0xd5/0x2b0
> [    9.389633]  ? blkdev_read_iter+0x6b/0x180
> [    9.389678]  ? lock_acquire+0xe5/0x2b0
> [    9.389720]  ? lock_is_held_type+0xcd/0x130
> [    9.389761]  ? find_held_lock+0x2b/0x80
> [    9.389813]  ? lock_acquired+0x1e9/0x3c0
> [    9.389864]  blkdev_read_iter+0x79/0x180
> [    9.389911]  ? local_clock_noinstr+0x17/0x110
> [    9.389975]  vfs_read+0x240/0x340
> [    9.390033]  ksys_read+0x61/0xd0
> [    9.390083]  do_syscall_64+0x74/0x3a0
> [    9.390143]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
> [    9.390204] RIP: 0033:0x7f6e048c5c5e
> [    9.390255] Code: 4d 89 d8 e8 34 bd 00 00 4c 8b 5d f8 41 8b 93 08 03
> 00 00 59 5e 48 83 f8 fc 74 11 c9 c3 0f 1f 80 00 00 00 00 48 8b 45 10 0f
> 05 <c9> c3 83 e2 39 83 fa 08 75 e7 e8 13 ff ff ff 0f 1f 00 f3 0f 1e fa
> [    9.390447] RSP: 002b:00007ffdf50cded0 EFLAGS: 00000202 ORIG_RAX:
> 0000000000000000
> [    9.390529] RAX: ffffffffffffffda RBX: 000000013aa55200 RCX:
> 00007f6e048c5c5e
> [    9.390619] RDX: 0000000000000200 RSI: 00007f6e0327f000 RDI:
> 0000000000000014
> [    9.390704] RBP: 00007ffdf50cdee0 R08: 0000000000000000 R09:
> 0000000000000000
> [    9.390789] R10: 0000000000000000 R11: 0000000000000202 R12:
> 0000000000000000
> [    9.390866] R13: 000055d5f99fa270 R14: 000055d5f96e5cb0 R15:
> 000055d5f96e5cc8
> [    9.390967]  </TASK>
> [    9.390997] irq event stamp: 54269
> [    9.391044] hardirqs last  enabled at (54279): [<ffffffff8138ac02>]
> __up_console_sem+0x52/0x60
> [    9.391141] hardirqs last disabled at (54288): [<ffffffff8138abe7>]
> __up_console_sem+0x37/0x60
> [    9.391240] softirqs last  enabled at (53796): [<ffffffff81304768>]
> irq_exit_rcu+0x78/0x110
> [    9.391327] softirqs last disabled at (53787): [<ffffffff81304768>]
> irq_exit_rcu+0x78/0x110
> [    9.391420] ---[ end trace 0000000000000000 ]---
>
> I haven't debugged this further yet, maybe you have an idea what
> could've caused it. On my side it's trivial to reproduce, so if you
> can't reproduce it just yell.

Hi Johannes,

Thanks for your report and for bisecting this.

A few questions:
1.  From the stack trace it looks like this is happening during a
block device read triggered by udev-worker's device probing. Does this
trigger for you consistently when the zoned device is probed or only
when running the btrfs xfstests generic/648?

2.  I tried to repro your setup by running:
sudo modprobe null_blk nr_devices=2 zoned=1 zone_size=256
zone_nr_conv=8 memory_backed=1
sudo mkdir -p /mnt/test && sudo mkdir -p /mnt/scratch
sudo mkfs.btrfs -f /dev/nullb0
sudo mount /dev/nullb0 /mnt/test
sudo ./check generic/648

with local.config set to:
TEST_DEV=/dev/nullb0
TEST_DIR=/mnt/test
SCRATCH_DEV=/dev/nullb1
SCRATCH_MNT=/mnt/scratch
export FSTYP=btrfs

but I’m not seeing the hang or the WARNING in dmesg show up.  I am
running it with PREEMPT(full) enabled. Does this match your setup or
are you doing something differently? Are you using null_blk or an
actual zoned device? If you are using null_blk, what module parameters
are you using?

3. If you're able to repro this consistently, would you be able to add
these lines right above the WARN_ON on line 487 and sharing what it
prints out?

+++ b/fs/iomap/buffered-io.c
@@ -484,6 +484,17 @@ static void iomap_read_end(struct folio *folio,
size_t bytes_submitted)
                 * to the IO helper, in which case we are responsible for
                 * unlocking the folio here.
                 */
+               if (bytes_submitted) {
+                       struct inode *inode = folio->mapping->host;
+                       struct block_device *bdev = inode->i_sb->s_bdev;
+
+                       pr_warn("bytes_submitted=%zu folio_size=%zu
blkbits=%u isize=%lld "
+                               "logical_bs=%u physical_bs=%u\n",
+                               bytes_submitted, folio_size(folio),
inode->i_blkbits,
+                               i_size_read(inode),
+                               bdev ? bdev_logical_block_size(bdev) : 0,
+                               bdev ? bdev_physical_block_size(bdev) : 0);
+               }
                WARN_ON_ONCE(bytes_submitted);
                folio_unlock(folio);
        }

Thanks,
Joanne
>
> Byte,
>
>      Johannes
>

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Johannes Thumshirn]

On 3/9/26 9:56 PM, Joanne Koong wrote:
> A few questions:
> 1.  From the stack trace it looks like this is happening during a
> block device read triggered by udev-worker's device probing. Does this
> trigger for you consistently when the zoned device is probed or only
> when running the btrfs xfstests generic/648?

Its only on g/648. I've also tried zoned XFS but g/648 didn't like my 
XFS and skipped the test case.


> 2.  I tried to repro your setup by running:
> sudo modprobe null_blk nr_devices=2 zoned=1 zone_size=256
> zone_nr_conv=8 memory_backed=1
> sudo mkdir -p /mnt/test && sudo mkdir -p /mnt/scratch
> sudo mkfs.btrfs -f /dev/nullb0
> sudo mount /dev/nullb0 /mnt/test
> sudo ./check generic/648
>
> with local.config set to:
> TEST_DEV=/dev/nullb0
> TEST_DIR=/mnt/test
> SCRATCH_DEV=/dev/nullb1
> SCRATCH_MNT=/mnt/scratch
> export FSTYP=btrfs
>
> but I’m not seeing the hang or the WARNING in dmesg show up.  I am
> running it with PREEMPT(full) enabled. Does this match your setup or
> are you doing something differently? Are you using null_blk or an
> actual zoned device? If you are using null_blk, what module parameters
> are you using?

I'm using a zloop device passed to a VM via virtio-blk. The zloop 
configuration is just:

echo "add id=0,zone_size_mb=256,conv_zones=4" > /dev/zloop-control

echo "add id=1,zone_size_mb=256,conv_zones=4" > /dev/zloop-control

The virtio-blk conifg is:

-drive driver=host_device,file=/dev/zloop0,if=virtio,cache.direct=on \

-drive driver=host_device,file=/dev/zloop1,if=virtio,cache.direct=on


The rootfs is on virtio-fs (using virtme-ng).

>
> 3. If you're able to repro this consistently, would you be able to add
> these lines right above the WARN_ON on line 487 and sharing what it
> prints out?
>
> +++ b/fs/iomap/buffered-io.c
> @@ -484,6 +484,17 @@ static void iomap_read_end(struct folio *folio,
> size_t bytes_submitted)
>                   * to the IO helper, in which case we are responsible for
>                   * unlocking the folio here.
>                   */
> +               if (bytes_submitted) {
> +                       struct inode *inode = folio->mapping->host;
> +                       struct block_device *bdev = inode->i_sb->s_bdev;
> +
> +                       pr_warn("bytes_submitted=%zu folio_size=%zu
> blkbits=%u isize=%lld "
> +                               "logical_bs=%u physical_bs=%u\n",
> +                               bytes_submitted, folio_size(folio),
> inode->i_blkbits,
> +                               i_size_read(inode),
> +                               bdev ? bdev_logical_block_size(bdev) : 0,
> +                               bdev ? bdev_physical_block_size(bdev) : 0);
> +               }
>                  WARN_ON_ONCE(bytes_submitted);
>                  folio_unlock(folio);
>          }

Here we go:

[   17.872952] bytes_submitted=1024 folio_size=4096 blkbits=12 
isize=5278880768 logical_bs=0 physical_bs=0

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Joanne Koong]

On Tue, Mar 10, 2026 at 4:44 AM Johannes Thumshirn
<Johannes.Thumshirn@wdc.com> wrote:
>
>
> >
> > 3. If you're able to repro this consistently, would you be able to add
> > these lines right above the WARN_ON on line 487 and sharing what it
> > prints out?
> >
> > +++ b/fs/iomap/buffered-io.c
> > @@ -484,6 +484,17 @@ static void iomap_read_end(struct folio *folio,
> > size_t bytes_submitted)
> >                   * to the IO helper, in which case we are responsible for
> >                   * unlocking the folio here.
> >                   */
> > +               if (bytes_submitted) {
> > +                       struct inode *inode = folio->mapping->host;
> > +                       struct block_device *bdev = inode->i_sb->s_bdev;
> > +
> > +                       pr_warn("bytes_submitted=%zu folio_size=%zu
> > blkbits=%u isize=%lld "
> > +                               "logical_bs=%u physical_bs=%u\n",
> > +                               bytes_submitted, folio_size(folio),
> > inode->i_blkbits,
> > +                               i_size_read(inode),
> > +                               bdev ? bdev_logical_block_size(bdev) : 0,
> > +                               bdev ? bdev_physical_block_size(bdev) : 0);
> > +               }
> >                  WARN_ON_ONCE(bytes_submitted);
> >                  folio_unlock(folio);
> >          }
>
> Here we go:
>
> [   17.872952] bytes_submitted=1024 folio_size=4096 blkbits=12
> isize=5278880768 logical_bs=0 physical_bs=0

Thank you, this is very helpful.

Is it correct that the block device's inode->i_blkbits doesn't reflect
its actual logical block size (512) or is that itself a bug somewhere
in the block layer? On the iomap side, it uses inode->i_blkbits to
determine whether or not an ifs should be attached and what logic to
correspondingly call. For this case, shouldn't the inode's i_blkbits
be 9 since "isize= 5278880768" indicates it's 512-byte aligned, not
4096-byte aligned? I'm not familiar with the block layer, so your
thoughts on this would be appreciated.

If it is fine for the block device's inode->i_blkbits to be a
different value from the logical block size, then I think the iomap
side needs this fix:

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index 867e8ac761c8..03a97361570f 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -543,7 +543,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
                         * helper, then the helper owns the folio and will end
                         * the read on it.
                         */
-                       if (*bytes_submitted == folio_len)
+                       if (*bytes_submitted == folio_len || !ifs)
                                ctx->cur_folio = NULL;
                }

Could you verify that this fixes the hang?

Thanks,
Joanne
>

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Joanne Koong]

On Tue, Mar 10, 2026 at 2:08 PM Joanne Koong <joannelkoong@gmail.com> wrote:
>
> On Tue, Mar 10, 2026 at 4:44 AM Johannes Thumshirn
> <Johannes.Thumshirn@wdc.com> wrote:
>
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index 867e8ac761c8..03a97361570f 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -543,7 +543,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
>                          * helper, then the helper owns the folio and will end
>                          * the read on it.
>                          */
> -                       if (*bytes_submitted == folio_len)
> +                       if (*bytes_submitted == folio_len || !ifs)
>                                 ctx->cur_folio = NULL;
>                 }

it should be this instead:

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index 867e8ac761c8..b803f518adaf 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -497,6 +497,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
        loff_t length = iomap_length(iter);
        struct folio *folio = ctx->cur_folio;
        size_t folio_len = folio_size(folio);
+       struct iomap_folio_state *ifs;
        size_t poff, plen;
        loff_t pos_diff;
        int ret;
@@ -508,7 +509,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
                return iomap_iter_advance(iter, length);
        }

-       ifs_alloc(iter->inode, folio, iter->flags);
+       ifs = ifs_alloc(iter->inode, folio, iter->flags);

        length = min_t(loff_t, length, folio_len - offset_in_folio(folio, pos));
        while (length) {
@@ -543,7 +544,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
                         * helper, then the helper owns the folio and will end
                         * the read on it.
                         */
-                       if (*bytes_submitted == folio_len)
+                       if (*bytes_submitted == folio_len || !ifs)
                                ctx->cur_folio = NULL;
                }

Thanks,
Joanne

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Darrick J. Wong]

On Tue, Mar 10, 2026 at 02:08:28PM -0700, Joanne Koong wrote:
> On Tue, Mar 10, 2026 at 4:44 AM Johannes Thumshirn
> <Johannes.Thumshirn@wdc.com> wrote:
> >
> >
> > >
> > > 3. If you're able to repro this consistently, would you be able to add
> > > these lines right above the WARN_ON on line 487 and sharing what it
> > > prints out?
> > >
> > > +++ b/fs/iomap/buffered-io.c
> > > @@ -484,6 +484,17 @@ static void iomap_read_end(struct folio *folio,
> > > size_t bytes_submitted)
> > >                   * to the IO helper, in which case we are responsible for
> > >                   * unlocking the folio here.
> > >                   */
> > > +               if (bytes_submitted) {
> > > +                       struct inode *inode = folio->mapping->host;
> > > +                       struct block_device *bdev = inode->i_sb->s_bdev;
> > > +
> > > +                       pr_warn("bytes_submitted=%zu folio_size=%zu
> > > blkbits=%u isize=%lld "
> > > +                               "logical_bs=%u physical_bs=%u\n",
> > > +                               bytes_submitted, folio_size(folio),
> > > inode->i_blkbits,
> > > +                               i_size_read(inode),
> > > +                               bdev ? bdev_logical_block_size(bdev) : 0,
> > > +                               bdev ? bdev_physical_block_size(bdev) : 0);
> > > +               }
> > >                  WARN_ON_ONCE(bytes_submitted);
> > >                  folio_unlock(folio);
> > >          }
> >
> > Here we go:
> >
> > [   17.872952] bytes_submitted=1024 folio_size=4096 blkbits=12
> > isize=5278880768 logical_bs=0 physical_bs=0
> 
> Thank you, this is very helpful.
> 
> Is it correct that the block device's inode->i_blkbits doesn't reflect
> its actual logical block size (512) or is that itself a bug somewhere
> in the block layer?

For a bdev, i_blkbits is only loosely tied to the storage device's block
size.  i_blkbits can't be smaller than log2(lba_size) and it can't be
larger than PAGE_SHIFT (or the max folio order if LBS is present).

Typically it starts out matching PAGE_SIZE, but a program that opens the
block device can call BLKBSZSET to set the block size.  This is usually
done by userspace filesystem tools to improve performance and/or ensure
that the kernel actually supports that block size.

Note that kernel filesystem drivers also call sb_set_blocksize to set
the block size.

The bdev block size seemingly resets to 4k after closing.

> On the iomap side, it uses inode->i_blkbits to
> determine whether or not an ifs should be attached and what logic to
> correspondingly call. For this case, shouldn't the inode's i_blkbits
> be 9 since "isize= 5278880768" indicates it's 512-byte aligned, not
> 4096-byte aligned? I'm not familiar with the block layer, so your
> thoughts on this would be appreciated.

<shrug> Regular files can have a larger i_blkbits than the underlying
storage device, so I don't see why bdevs would be different?

--D

> If it is fine for the block device's inode->i_blkbits to be a
> different value from the logical block size, then I think the iomap
> side needs this fix:
> 
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index 867e8ac761c8..03a97361570f 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -543,7 +543,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
>                          * helper, then the helper owns the folio and will end
>                          * the read on it.
>                          */
> -                       if (*bytes_submitted == folio_len)
> +                       if (*bytes_submitted == folio_len || !ifs)
>                                 ctx->cur_folio = NULL;
>                 }
> 
> Could you verify that this fixes the hang?
> 
> Thanks,
> Joanne
> >
> 

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Johannes Thumshirn]

On 3/10/26 10:55 PM, Joanne Koong wrote:
> On Tue, Mar 10, 2026 at 2:08 PM Joanne Koong <joannelkoong@gmail.com> wrote:
>> On Tue, Mar 10, 2026 at 4:44 AM Johannes Thumshirn
>> <Johannes.Thumshirn@wdc.com> wrote:
>>
>> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
>> index 867e8ac761c8..03a97361570f 100644
>> --- a/fs/iomap/buffered-io.c
>> +++ b/fs/iomap/buffered-io.c
>> @@ -543,7 +543,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
>>                           * helper, then the helper owns the folio and will end
>>                           * the read on it.
>>                           */
>> -                       if (*bytes_submitted == folio_len)
>> +                       if (*bytes_submitted == folio_len || !ifs)
>>                                  ctx->cur_folio = NULL;
>>                  }
> it should be this instead:
>
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index 867e8ac761c8..b803f518adaf 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -497,6 +497,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
>          loff_t length = iomap_length(iter);
>          struct folio *folio = ctx->cur_folio;
>          size_t folio_len = folio_size(folio);
> +       struct iomap_folio_state *ifs;
>          size_t poff, plen;
>          loff_t pos_diff;
>          int ret;
> @@ -508,7 +509,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
>                  return iomap_iter_advance(iter, length);
>          }
>
> -       ifs_alloc(iter->inode, folio, iter->flags);
> +       ifs = ifs_alloc(iter->inode, folio, iter->flags);
>
>          length = min_t(loff_t, length, folio_len - offset_in_folio(folio, pos));
>          while (length) {
> @@ -543,7 +544,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
>                           * helper, then the helper owns the folio and will end
>                           * the read on it.
>                           */
> -                       if (*bytes_submitted == folio_len)
> +                       if (*bytes_submitted == folio_len || !ifs)
>                                  ctx->cur_folio = NULL;
>                  }
>
> Thanks,
> Joanne
>
This version is making the test pass again,

Reported-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>

Tested-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>

Thanks a lot for this!

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Joanne Koong]

On Tue, Mar 10, 2026 at 2:59 PM Darrick J. Wong <djwong@kernel.org> wrote:
>
> On Tue, Mar 10, 2026 at 02:08:28PM -0700, Joanne Koong wrote:
> > On Tue, Mar 10, 2026 at 4:44 AM Johannes Thumshirn
> > <Johannes.Thumshirn@wdc.com> wrote:
> > >
> > >
> > > >
> > > > 3. If you're able to repro this consistently, would you be able to add
> > > > these lines right above the WARN_ON on line 487 and sharing what it
> > > > prints out?
> > > >
> > > > +++ b/fs/iomap/buffered-io.c
> > > > @@ -484,6 +484,17 @@ static void iomap_read_end(struct folio *folio,
> > > > size_t bytes_submitted)
> > > >                   * to the IO helper, in which case we are responsible for
> > > >                   * unlocking the folio here.
> > > >                   */
> > > > +               if (bytes_submitted) {
> > > > +                       struct inode *inode = folio->mapping->host;
> > > > +                       struct block_device *bdev = inode->i_sb->s_bdev;
> > > > +
> > > > +                       pr_warn("bytes_submitted=%zu folio_size=%zu
> > > > blkbits=%u isize=%lld "
> > > > +                               "logical_bs=%u physical_bs=%u\n",
> > > > +                               bytes_submitted, folio_size(folio),
> > > > inode->i_blkbits,
> > > > +                               i_size_read(inode),
> > > > +                               bdev ? bdev_logical_block_size(bdev) : 0,
> > > > +                               bdev ? bdev_physical_block_size(bdev) : 0);
> > > > +               }
> > > >                  WARN_ON_ONCE(bytes_submitted);
> > > >                  folio_unlock(folio);
> > > >          }
> > >
> > > Here we go:
> > >
> > > [   17.872952] bytes_submitted=1024 folio_size=4096 blkbits=12
> > > isize=5278880768 logical_bs=0 physical_bs=0
> >
> > Thank you, this is very helpful.
> >
> > Is it correct that the block device's inode->i_blkbits doesn't reflect
> > its actual logical block size (512) or is that itself a bug somewhere
> > in the block layer?
>
> For a bdev, i_blkbits is only loosely tied to the storage device's block
> size.  i_blkbits can't be smaller than log2(lba_size) and it can't be
> larger than PAGE_SHIFT (or the max folio order if LBS is present).
>
> Typically it starts out matching PAGE_SIZE, but a program that opens the
> block device can call BLKBSZSET to set the block size.  This is usually
> done by userspace filesystem tools to improve performance and/or ensure
> that the kernel actually supports that block size.
>
> Note that kernel filesystem drivers also call sb_set_blocksize to set
> the block size.
>
> The bdev block size seemingly resets to 4k after closing.

Thanks for the explanation. I hadn't realized inode->i_blkbits is only
loosely tied to the actual I/O granularity.

Thanks,
Joanne
>
> > On the iomap side, it uses inode->i_blkbits to
> > determine whether or not an ifs should be attached and what logic to
> > correspondingly call. For this case, shouldn't the inode's i_blkbits
> > be 9 since "isize= 5278880768" indicates it's 512-byte aligned, not
> > 4096-byte aligned? I'm not familiar with the block layer, so your
> > thoughts on this would be appreciated.
>
> <shrug> Regular files can have a larger i_blkbits than the underlying
> storage device, so I don't see why bdevs would be different?
>
> --D
>

Re: Hang in generic/648 on zoned btrfs after aa35dd5cbc06 ("iomap: fix invalid folio access after folio_end_read()")

[Joanne Koong]

On Tue, Mar 17, 2026 at 2:35 AM Johannes Thumshirn
<Johannes.Thumshirn@wdc.com> wrote:
>
> On 3/10/26 10:55 PM, Joanne Koong wrote:
> > On Tue, Mar 10, 2026 at 2:08 PM Joanne Koong <joannelkoong@gmail.com> wrote:
> >> On Tue, Mar 10, 2026 at 4:44 AM Johannes Thumshirn
> >> <Johannes.Thumshirn@wdc.com> wrote:
> >>
> >> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> >> index 867e8ac761c8..03a97361570f 100644
> >> --- a/fs/iomap/buffered-io.c
> >> +++ b/fs/iomap/buffered-io.c
> >> @@ -543,7 +543,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
> >>                           * helper, then the helper owns the folio and will end
> >>                           * the read on it.
> >>                           */
> >> -                       if (*bytes_submitted == folio_len)
> >> +                       if (*bytes_submitted == folio_len || !ifs)
> >>                                  ctx->cur_folio = NULL;
> >>                  }
> > it should be this instead:
> >
> > diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> > index 867e8ac761c8..b803f518adaf 100644
> > --- a/fs/iomap/buffered-io.c
> > +++ b/fs/iomap/buffered-io.c
> > @@ -497,6 +497,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
> >          loff_t length = iomap_length(iter);
> >          struct folio *folio = ctx->cur_folio;
> >          size_t folio_len = folio_size(folio);
> > +       struct iomap_folio_state *ifs;
> >          size_t poff, plen;
> >          loff_t pos_diff;
> >          int ret;
> > @@ -508,7 +509,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
> >                  return iomap_iter_advance(iter, length);
> >          }
> >
> > -       ifs_alloc(iter->inode, folio, iter->flags);
> > +       ifs = ifs_alloc(iter->inode, folio, iter->flags);
> >
> >          length = min_t(loff_t, length, folio_len - offset_in_folio(folio, pos));
> >          while (length) {
> > @@ -543,7 +544,7 @@ static int iomap_read_folio_iter(struct iomap_iter *iter,
> >                           * helper, then the helper owns the folio and will end
> >                           * the read on it.
> >                           */
> > -                       if (*bytes_submitted == folio_len)
> > +                       if (*bytes_submitted == folio_len || !ifs)
> >                                  ctx->cur_folio = NULL;
> >                  }
> >
> > Thanks,
> > Joanne
> >
> This version is making the test pass again,
>
> Reported-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
>
> Tested-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
>
> Thanks a lot for this!
>

Thanks for reporting this and testing / helping to debug the issue.
I'll submit the fix for this upstream today.

Thanks,
Joanne