On Monday, 27 March 2023 15:59:37 CEST Marcel Camijn wrote: > We recently got new hardware with a new raid controller but noticed that the > required module MPI3MR is not enabled. The latest uploaded kernel image has > "#CONFIG_SCSI_MPI3MR is not set" in the config. > > The module seems to be included upstream since 5.14. > > For current stable we can work around this, but for bookworm is would be > nice to have it enabled. I have created a branch to enable it and when Salsa's CI succeeds (enough) I can turn that into a MR. While looking into it, I did find some troubling aspects though. I noticed that in upstream's master branch there were quite some commits *after* 6.1 of which a reasonable amount has been backported to 6.1. For example, in 6.1.21 there are FIVE commits which fix a memory leak. Those got backported because the original commits contained "Fixes: <commit ID>" There are however also some commits which do NOT have the "Fixes:" tag, like: https://lore.kernel.org/all/20230228140835.4075-7-ranjan.kumar@broadcom.com/ titled: "mpi3mr: Bad drive in topology results kernel crash" There are some more and for that compare upstream's master branch with 6.1 and look for the ones which do not have the "Fixes:" tag. Unless explicit action is taken to request a backport, f.e. the above mentioned commit will not get backported to 6.1 ... which could lead to a kernel crash. I will propose the MR, but I also think the Debian kernel maintainers should be aware of this which could result in the decision to NOT merge the MR. Keeping an eye on and involving with upstream wrt those (potential) issues seems like a good strategy. HTH, Diederik
Attachment:
signature.asc
Description: This is a digitally signed message part.