Hello,
the lpfc boot regression reported in this bug is now fixed upstream in
Linux 6.1.191:
commit 0da9abd8ef9da257c947d0d49bff4b6781d1f92d
scsi: lpfc: Fix use-after-free of mboxq in
lpfc_sli4_driver_resource_setup()
Fixes: 299f5baa1321 ("scsi: lpfc: Fix memory leak in
lpfc_sli4_driver_resource_setup()")
https://git.kernel.org/stable/c/0da9abd8ef9da257c947d0d49bff4b6781d1f92d
The fix moves mempool_free(mboxq) after the error handling of
lpfc_get_sli4_parameters(), as in mainline and the other stable trees.
It is a two-line change and applies cleanly on top of 6.1.190.
On systems with an Emulex LPe35000/LPe36000 HBA
(10df:f400, firmware 14.4.730.15) this kernel crashes with slab
corruption while lpfc probes and does not boot. Affected hosts therefore
have to stay on 6.1.170-1 (6.1.0-45).
Could you please include this fix in a regression update for
bookworm-security (e.g. 6.1.190-2 / DLA-4841-2), or ship 6.1.191 soon?
Thanks
