On Wed Dec 04 19, Lu Baolu wrote:
>Hi,
>
>On 12/3/19 5:56 PM, Jerry Snitselaar wrote:
>>On Tue Dec 03 19, Lu Baolu wrote:
>>>Hi,
>>>
>>>On 12/3/19 12:13 AM, Jerry Snitselaar wrote:
>>>>On Mon Dec 02 19, Jerry Snitselaar wrote:
>>>>>On Mon Dec 02 19, Lu Baolu wrote:
>>>>>>Hi,
>>>>>>
>>>>>>On 12/2/19 2:34 PM, Jerry Snitselaar wrote:
>>>>>>>We are seeing DMAR PTE read access not set errors when booting a
>>>>>>>kernel with default passthrough, both with a test kernel and with
>>>>>>>a 5.4.0 kernel. Previously we would see a number of identity mappings
>>>>>>>being set related to the rmrrs, and now they aren't seen and we get
>>>>>>>the dmar pte errors as devices touch those regions. From 
>>>>>>>what I can tell
>>>>>>>currently df4f3c603aeb ("iommu/vt-d: Remove static 
>>>>>>>identity map code")
>>>>>>>removed the bit of code in init_dmars that used to set up those
>>>>>>>mappings:
>>>>>>>
>>>>>>>-       /*
>>>>>>>-        * For each rmrr
>>>>>>>-        *   for each dev attached to rmrr
>>>>>>>-        *   do
>>>>>>>-        *     locate drhd for dev, alloc domain for dev
>>>>>>>-        *     allocate free domain
>>>>>>>-        *     allocate page table entries for rmrr
>>>>>>>-        *     if context not allocated for bus
>>>>>>>-        *           allocate and init context
>>>>>>>-        *           set present in root table for this bus
>>>>>>>-        *     init context with domain, translation etc
>>>>>>>-        *    endfor
>>>>>>>-        * endfor
>>>>>>>-        */
>>>>>>>-       pr_info("Setting RMRR:\n");
>>>>>>>-       for_each_rmrr_units(rmrr) {
>>>>>>>-               /* some BIOS lists non-exist devices in 
>>>>>>>DMAR table. */
>>>>>>>-               for_each_active_dev_scope(rmrr->devices, 
>>>>>>>rmrr->devices_cnt,
>>>>>>>-                                         i, dev) {
>>>>>>>-                       ret = iommu_prepare_rmrr_dev(rmrr, dev);
>>>>>>>-                       if (ret)
>>>>>>>-                               pr_err("Mapping reserved 
>>>>>>>region failed\n");
>>>>>>>-               }
>>>>>>>-       }
>>>>>>>
>>>>>>>si_domain_init now has code that sets identity maps for 
>>>>>>>devices in rmrrs, but
>>>>>>>only for certain devices.
>>>>>>
>>>>>>On which device, are you seeing this error? Is it a rmrr 
>>>>>>locked device?
>>>>>>
>>>>>>Best regards,
>>>>>>baolu
>>>>>>
>>>>>
>>>>>Almost all of the messages are for the ilo, but there also is 
>>>>>a message for
>>>>>the smart array raid bus controller.
>>>>>
>>>>
>>>>Also seeing it with a dl380 gen9 system, where the raid bus controller
>>>>is getting the error.
>>>
>>>Does it help if you remove
>>>
>>>                       if (device_is_rmrr_locked(dev))
>>>                               continue;
>>>
>>>in si_domain_init()?
>>>
>>
>>Unfortunately it still spits out a bunch of error messages.
>
>Can you please show me the whole dmesg?
>
>Best regards,
>-baolu
>

Attaching console output (can't get to a point to actually log in) and
config that is used to build that kernel.