https://bugzilla.kernel.org/show_bug.cgi?id=215906 --- Comment #22 from Michał Pecio (michal.pecio@xxxxxxxxx) --- That's good to hear. Just to be sure, can you confirm that without this quirk the problem still exists today, on the same machine running the same kernel? This enables the XHCI_ZHAOXIN_TRB_FETCH quirk which allocates 8KB for each 4KB transfer ring segment and leaves the upper 4KB empty, so that DMA accesses beyond the 4KB segment end up hitting valid and harmless memory. It wastes a few KB of memory per endpoint, and RasPI came up with a more efficient solution, but this is something already implemented here so we could try it immediately without patching. BTW, there is speculation that the Zhaoxin bug may be caused by odd page size requirements of some controllers. If your kernel has debugfs enabled, please also post the output of: grep PAGESIZE /sys/kernel/debug/usb/xhci/*/reg-op lspci -nn |grep USB -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug.
