Re: [PATCH v8 0/9] Ext4 Fast Commit Performance Patchset

["Theodore Ts'o" <tytso@mit.edu>]

On Thu, May 01, 2025 at 12:14:04AM -0400, Theodore Ts'o wrote:
> 
> I was applying your patches on 6.15-rc3 and this evening I tried
> applying them against 6.15-rc2, which has been failing as well.
> 
> There's nothing else on the ext4 dev branch yet, but there were a
> large number of ext4 patches which landed between 6.14-rc2 and
> 6.15-rc2.

I tried bisecting your patch series using:

  kvm-xfstests -c ext4/fast_commit generic/127 generic/241 generic/418

and the first bad commit was "ext4: rework fast commit commit path".

    	      	  	     	    	   - Ted

git bisect start
# status: waiting for both good and bad commits
# good: [9c32cda43eb78f78c73aee4aa344b777714e259b] Linux 6.15-rc3
git bisect good 9c32cda43eb78f78c73aee4aa344b777714e259b
# status: waiting for bad commit, 1 good commit known
# bad: [a147d000c6914a51becf0d32bbb8c9124e45f6ed] ext4: hold s_fc_lock while during fast commit
git bisect bad a147d000c6914a51becf0d32bbb8c9124e45f6ed
# good: [8ffd015db85fea3e15a77027fda6c02ced4d2444] Linux 6.15-rc2
git bisect good 8ffd015db85fea3e15a77027fda6c02ced4d2444
# bad: [bfd1ce278bb3cfea0ad7017de608691a83c372bf] ext4: rework fast commit commit path
git bisect bad bfd1ce278bb3cfea0ad7017de608691a83c372bf
# good: [36eb0b696b73895b7e40dbf3043275fdc9bcb1fe] ext4: for committing inode, make ext4_fc_track_inode wait
git bisect good 36eb0b696b73895b7e40dbf3043275fdc9bcb1fe
# good: [94cee94db3acfe189463f4343ca7bd9c83570cb7] ext4: mark inode dirty before grabbing i_data_sem in ext4_setattr
git bisect good 94cee94db3acfe189463f4343ca7bd9c83570cb7
# first bad commit: [bfd1ce278bb3cfea0ad7017de608691a83c372bf] ext4: rework fast commit commit path