Re: 6.12 WARNING in netfs_consume_read_data()

[Max Kellermann]

On Fri, Nov 29, 2024 at 11:29 PM Max Kellermann
<max.kellermann@xxxxxxxxx> wrote:
> I have encountered multiple hangs with that branch; for example:

Similar hangs wth 6.12.2 (vanilla without your "netfs-writeback" branch):

[<0>] folio_wait_bit_common+0x23a/0x4f0
[<0>] folio_wait_private_2+0x37/0x70
[<0>] netfs_invalidate_folio+0x168/0x520
[<0>] truncate_cleanup_folio+0x281/0x340
[<0>] truncate_inode_pages_range+0x1bb/0x780
[<0>] ceph_evict_inode+0x17e/0x6b0
[<0>] evict+0x331/0x780
[<0>] __dentry_kill+0x17b/0x4f0
[<0>] dput+0x2a6/0x4a0
[<0>] __fput+0x36d/0x910
[<0>] __x64_sys_close+0x78/0xd0
[<0>] do_syscall_64+0x64/0x100
[<0>] entry_SYSCALL_64_after_hwframe+0x76/0x7e

netfs in 6.12 remains pretty broken, both with and without your fixes.

The sad thing is that 6.11 has just been EOLed. Linux 6.6 is the only
sane kernel.