ch348 serial failling on linux-next

[Corentin LABBE <clabbe@baylibre.com>]

Hello

I am working on mainlining ch348 driver and I have some CI test to verify it works.
The ch348 is an octo port chip, and for testing it I did the following setup:
port0 to a FTDI, port1 to a pl2303, port2 to another FTDI, port 6 and 7 linked together.
For each couple, I send/recv data for a list of baudrate and compare results.

My test suite was okay on stable, 6.13-rc6, but when testing linux-next it fail since 20250110.
And only on port0<->FTDI.
I bisected the issue to: libfs: Use d_children list to iterate simple_offset directories
This is totally unrelated, but the fail is not random and persistant.
I tried also next-20250113, next-20250114, same fail
In the mean time, 6.13-rc7 came out and works.

What tool I use to detect it ?
I use https://github.com/montjoie/lava-tests/blob/master/test2a2.py
./test2a2 --port0 /dev/ttyUSB1 --port1 /dev/ttyUSB0
serial test v0 for /dev/ttyUSB1 /dev/ttyUSB0
DEBUG: serial test /dev/ttyUSB1 /dev/ttyUSB0
TRY /dev/ttyUSB1 to /dev/ttyUSB0 baud=9600
DEBUG: send_recv 32
DEBUG: sent 4/32 sent=4 remains=28
DEBUG: RECV 0
DEBUG: sent 8/32 sent=12 remains=20
DEBUG: RECV 0
DEBUG: sent 14/32 sent=26 remains=6
DEBUG: RECV 0
DEBUG: sent 6/32 sent=32 remains=0
DEBUG: RECV 0
DEBUG: RECV 0
DEBUG: RECV 0
DEBUG: RECV 0
DEBUG: RECV 0
DEBUG: RECV 0
DEBUG: RECV 0
DEBUG: RECV 0
DEBUG: string are different

So /dev/ttyUSB1 is sending data to /dev/ttyUSB0 and the latter never received it.
Same problem for the opposite, /dev/ttyUSB0 to /dev/ttyUSB1

Nothing could be found in dmesg, no clue.

Okay now, I dont know how to go further, the bisect result is too unrelated to the problem.

I have retried a second bisect from a different start/end
git bisect start
# status : en attente d'un commit bon et d'un commit mauvais
# bad: [0907e7fb35756464aa34c35d6abb02998418164b] Add linux-next specific files for 20250117
git bisect bad 0907e7fb35756464aa34c35d6abb02998418164b
# status : en attente de bon(s) commit(s), un mauvais commit connu
# good: [5bc55a333a2f7316b58edc7573e8e893f7acb532] Linux 6.13-rc7
git bisect good 5bc55a333a2f7316b58edc7573e8e893f7acb532
# bad: [195cedf4deacf84167c32b866ceac1cf4a16df15] Merge branch 'main' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git
git bisect bad 195cedf4deacf84167c32b866ceac1cf4a16df15
# good: [e8c0711b153b0db806410d8b31ed23b590f4eab4] Merge branch 'xtensa-for-next' of git://github.com/jcmvbkbc/linux-xtensa.git
git bisect good e8c0711b153b0db806410d8b31ed23b590f4eab4
# bad: [be4d7a3e7815249ca857f618dee81549f745cdbc] Merge branch 'docs-next' of git://git.lwn.net/linux.git
git bisect bad be4d7a3e7815249ca857f618dee81549f745cdbc
# good: [19096ecb142b72cebf03d8316c1d96192620e23a] Merge branch 'master' of https://github.com/Paragon-Software-Group/linux-ntfs3.git
git bisect good 19096ecb142b72cebf03d8316c1d96192620e23a
# bad: [51af4bb6edb606a7d1160323d65e6715969124e1] Merge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux.git
git bisect bad 51af4bb6edb606a7d1160323d65e6715969124e1
# bad: [9ce2f898c13763037516269044d4658a2eabde61] Merge branch 'vfs-6.14.libfs' into vfs.all
git bisect bad 9ce2f898c13763037516269044d4658a2eabde61
# good: [3be5a57e3e36d3c2b6532a2262472199da972407] Merge branch 'vfs-6.14.misc' into vfs.all
git bisect good 3be5a57e3e36d3c2b6532a2262472199da972407
# good: [fda429aeb9f70b1f4f3b63d80f40e442a24f985a] Merge branch 'kernel-6.14.cred' into vfs.all
git bisect good fda429aeb9f70b1f4f3b63d80f40e442a24f985a
# good: [3ab8a0b2a0ff1038412cd644b51714e35970f415] selftests: add listmount() iteration tests
git bisect good 3ab8a0b2a0ff1038412cd644b51714e35970f415
# good: [5f677209c2642cf289867ca86f65a04c47265109] Merge branch 'vfs-6.14.mount' into vfs.all
git bisect good 5f677209c2642cf289867ca86f65a04c47265109
# good: [b662d858131da9a8a14e68661656989b14dbf113] Revert "libfs: fix infinite directory reads for offset dir"
git bisect good b662d858131da9a8a14e68661656989b14dbf113
# bad: [b9b588f22a0c049a14885399e27625635ae6ef91] libfs: Use d_children list to iterate simple_offset directories
git bisect bad b9b588f22a0c049a14885399e27625635ae6ef91
# good: [68a3a65003145644efcbb651e91db249ccd96281] libfs: Replace simple_offset end-of-directory detection
git bisect good 68a3a65003145644efcbb651e91db249ccd96281
# first bad commit: [b9b588f22a0c049a14885399e27625635ae6ef91] libfs: Use d_children list to iterate simple_offset directories

Since 2 different bisect point the same commit, and that the issue is not random, I am confident the bad commit is the real one, but why ?

I am trying to create a second setup to exclude hardware problem, but hardware will fail always and not only with some commit present.

I am a bit lost

Regards