Hi Martin,
On 13/06/23 22:57, Martin Steigerwald wrote:
Michael Schmitz - 13.06.23, 10:18:24 CEST:
Am 13.06.2023 um 19:25 schrieb Martin Steigerwald:
Hi Michael, hi Jens, Hi Geert.
Michael Schmitz - 22.08.22, 22:56:10 CEST:
On 23/08/22 08:41, Jens Axboe wrote:
On 8/22/22 2:39 PM, Michael Schmitz wrote:
Hi Jens,
will do - just waiting to hear back what needs to be done
regarding
backporting issues raised by Geert.
It needs to go upstream first before it can go to stable. Just
mark
it with the right Fixes tags and that will happen automatically.
[…]
thanks - the Fixes tag in my patches refers to Martin's bug report
and won't be useful to decide how far back this must be applied.
Now the bug pre-dates git, making the commit to 'fix'
1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 ... That one's a bit
special, please yell if you want me to lie about this and use a
later commit specific to the partition parser code.
After this discussion happened I thought the patch went in. However…
as John Paul Adrian asked in "Status of affs support in the kernel
and affstools" thread on linux-m68k and debian-68k mailing list, I
searched for the patch in git history but did not find it.
I may have messed that one up, as it turns out. Last version was v9
which I had to resend twice, and depending on what Jens uses to keep
track of patches, the resends may not have shown up in his tool. I
should have bumped the version number instead.
I'll see if my latest version still applies cleanly ...
Many thanks!
Would be nice to see it finally go in.
My last version (v9) still applies, but that one still threw a sparse
warning for patch 2:
Link:https://lore.kernel.org/all/202208231319.Ng5RTzzg-lkp@xxxxxxxxx
Not sure how to treat that one - rdb_CylBlocks is not declared as big
endian so the warning is correct, but as far as I can see, for all
practical purposes rdb_CylBlocks would be expected to be in big endian
order (partition usually prepared on a big endian system)?
I can drop the be32_to_cpu conversion (and would expect to see a warning
printed on little endian systems), or force the cast to __be32. Or
rather drop that consistency check outright...
Cheers,
Michael
Best,
