On 6/25/19 8:42 AM, Christoph Hellwig wrote: >> Do you think we could test a kernel which just sets the flag unconditionally >> to see whether this is the problem that causes the issues on these machines? > > Might be worth a test. Do you know what device failed? Might be one > with a dma mask < 32-bit? I can reproduce the crash when trying to load the module for the USB controllers, for example. Loading the kernel module for the SATA controllers provokes the backtrace as well. I have skimmed through the code a bit , but I'm not sure whether I understand the code in kernel/dma/direct.c correctly, so my suggestion would be to just set GFP_DMA32 in __dma_direct_alloc_pages() unconditionally for a test. Would that be enough for a test? Adrian -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaubitz@xxxxxxxxxx `. `' Freie Universitaet Berlin - glaubitz@xxxxxxxxxxxxxxxxxxx `- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
|