On 27.11.2019 16:18, Durrant, Paul wrote: >> -----Original Message----- >> From: Tian, Kevin <kevin.t...@intel.com> >> Sent: 25 November 2019 08:22 >> To: Durrant, Paul <pdurr...@amazon.com>; xen-devel@lists.xenproject.org >> Cc: Jan Beulich <jbeul...@suse.com>; Andrew Cooper >> <andrew.coop...@citrix.com>; Wei Liu <w...@xen.org>; Roger Pau Monné >> <roger....@citrix.com> >> Subject: RE: [PATCH] x86 / iommu: set up a scratch page in the quarantine >> domain >> >>> From: Paul Durrant [mailto:pdurr...@amazon.com] >>> Sent: Wednesday, November 20, 2019 8:09 PM >>> >>> This patch introduces a new iommu_op to facilitate a per-implementation >>> quarantine set up, and then further code for x86 implementations >>> (amd and vtd) to set up a read/wrote scratch page to serve as the >> source/ >>> target for all DMA whilst a device is assigned to dom_io. >>> >>> The reason for doing this is that some hardware may continue to re-try >>> DMA, despite FLR, in the event of an error. Having a scratch page mapped >>> will allow pending DMA to drain and thus quiesce such buggy hardware. >> >> then there is no diagnostics at all since all faults are quiescent now... >> why do we want to support such buggy hardware? Is it better to make >> it an default-off option since buggy is supposed to niche case? > > I guess it could be a command line option... perhaps making the new > 'iommu=quarantine' boolean into something more complex, but I'm not > sure it's really worth it. Perhaps a compile time option would be > better?
Yet another option: How about installing the scratch page mappings only after a (handful of) IOMMU faults? But of course there was the related earlier question of whether indeed our turning off of bus mastering doesn't already help silencing the faults. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
