On 03/29/22 at 11:17am, Simon Horman wrote: > On Sat, Mar 26, 2022 at 02:55:43PM +0800, RuiRui Yang wrote: > > Hi Simon, > > > > Recently RH CKI detected a kexec/kdump failure with kexec_load being > > used, the 2nd kernel can not boot, just reset to bios/firmware > > instead. > > > > This only happens with gcc 12.x compiled kexec-tools, add > > "-fno-tree-vectorize" to purgatory/Makefile CFLAGS make the bug > > dissappear, but I hope people can help to verify and confirm it does > > fix the problem. It would be better to wait a few days so that we > > can fix it in the coming release, what do you think? > > > > Cced Baoquan who is handling the issue.. > > > > Thanks! > > Hi RuiRui, > > thanks for letting us know. > > I'm happy with delaying the release for a short time to accommodate > resolving this problem. Please let us know when there is some progress. Thanks for quick response, Simon. I have sent a patch to fix this, please check the patch v2 with subject 'purgatory: do not enable vectorization automatically for purgatory compiling'. I tested on machine where the issue can stably reproduced, and can confirm adding -fno-tree-vectorize to compiling options of purgatory can fix the issue. Dave, please check if it is expected from you. _______________________________________________ kexec mailing list kexec@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/kexec
