On Tue, 15 Apr 2025, Jan Kara wrote:
On Mon 14-04-25 18:36:41, Davidlohr Bueso wrote:
On Wed, 09 Apr 2025, Luis Chamberlain wrote:
> corruption can still happen even with the spin lock held. A test was
> done using vanilla Linux and adding a udelay(2000) right before we
> spin_lock(&bd_mapping->i_private_lock) on __find_get_block_slow() and
> we can reproduce the same exact filesystem corruption issues as observed
> without the spinlock with generic/750 [1].
fyi I was actually able to trigger this on a vanilla 6.15-rc1 kernel,
not even having to add the artificial delay.
OK, so this is using generic/750, isn't it? How long did you have to run it
to trigger this? Because I've never seen this tripping...
Correct, generic/750. It triggered after 90+ hrs running.
