On Mar 04, 2025 / 10:25, Milan Broz wrote: > On 2/28/25 3:32 PM, Ondrej Kozina wrote: > > On 28/02/2025 15:08, Jens Axboe wrote: > > > On 2/28/25 4:59 AM, Ondrej Kozina wrote: > > > > Hi Jens, > > > > > > > > this patch introduced regression to locked SED OPAL2 devices. The locked region no longer returns -EIO upon IO. On read the caller receives block of zeroes, on write it does not report any error either. In both cases, previous to this patch, the caller would get IO error (expected) with locked device. > > > > > > > > It was discovered by cryptsetup testsuite specifically https://gitlab.com/cryptsetup/cryptsetup/-/blob/main/tests/compat-test-opal > > > > > > > > I've attached a simple patch that changes the ioerror condition and it fixed the problem with SED OPAL2 devices for me. > > > > > > > > #regzbot introduced: 1f47ed294a2bd577d5ae43e6e28e1c9a3be4a833 > > > > > > Oops thanks - does someone want to send a "real" patch with commit message > > > etc, or do you just want me to queue something up? > > > > > > > To be honest, my patch was just a quick hack and I'm not sure if it is > > the correct one in general. But I will test (and report) a fix when it > > lands here. > > Hi Jens, > > do you have some fix for this issue anywhere? > > We can easily test the patch (at least for Opal) as our CI is currently > completely red because of this issue :) > Thanks, > Milan > I created fix candidate patches to address the blktests nvme/039 failure [1]. This may work for the failures Ondrej and Milan observe too, hopefully. Jens, Alan, could you take a look in the patches and see if they make sense? [1] https://lkml.kernel.org/linux-block/20250311024144.1762333-1-shinichiro.kawasaki@xxxxxxx/
