Hi Jens,

We (Fedora) just got this bug report: https://bugzilla.redhat.com/show_bug.cgi?id=1729678

But this is not Fedora specific and as the Fedora bug report mentioned has already been reported in bugzilla.kernel.org, but it seems that SATA bugs reported in bugzilla.kernel.org still end up getting assigned to Tejun...

Hence this email to bring the 2 upstream bugs to your attention: https://bugzilla.kernel.org/show_bug.cgi?id=201693 https://bugzilla.kernel.org/show_bug.cgi?id=203475

The second bug report has a patch which disables queued TRIM support on the 860, like it is already done for the 850.

This fixes things for the second reporter but I'm not sure if that is enough, since in the first bug report the reporter mentions trying "libata.force=noncqtrim" not helping and I've verified his dmesg and the parameter is being passed and processed correctly ultimately resulting in:

ata6.00: ATA-11: Samsung SSD 860 EVO 2TB, RVT01B6Q, max UDMA/133 ata6.00: disabling queued TRIM support

The 2 reporters do have different sata chipsets though, so maybe for the "ASMedia ASM1062 Serial ATA Controller" from the second bug report this is enough, but there is some other issue when combined with the SATA controller in AMD socket AM3 mobo chipsets?

The first bug report talks about FPDMA errors, so maybe the AM3 mobos need ATA_HORKAGE_BROKEN_FPDMA_AA?

Still it seems like a good idea to merge the patch from the 2nd bug since it at least helps some people (there is another using confirming the patch helps in the bug).

Regards,

Hans