On 9/12/18 1:39 PM, wrote: > On Mon, 2018-09-10 at 19:00 +0300, Sergey Miroshnichenko wrote: >> >> Yes, missing a real EEH event is possible, unfortunately, and it is >> indeed worth mentioning. >> >> To reduce this probability the next patchset I'll post in a few days >> among other things puts all the affected device drivers to pause during >> rescan, mainly because of moving BARs and bridge windows, but it will >> also help here a bit. > > How do you deal with moving BARs etc... within the segmenting > restrictions of EEH ? >
Actually, [un]fortunately, we haven't encountered any segmenting issues yet, but to move BARs we are using the same existing mechanism in Linux kernel that re-enumerated the PCIe topology during startup with "pci=realloc" and PCI_REASSIGN_ALL_BUS. What restrictions must be broken to provoke a segmenting event? Are there any other limitations on segmenting besides keeping all the BARs of the PHB within its huge M32+M64 segments which are 2GiB+4GiB on our setup? > It's a horrible mess right now and I don't know if the current code can > even work properly to be honest. > > Cheers, > Ben. > > Best regards, Serge