[Crash-utility] Re: [PATCH 2/2] Fix "kmem [-s]" command on Linux 7.1 and later

Hi Dave, thank you for testing the patches. On 2026/08/10 19:39, Dave Young wrote: > Kairui preferred to use the gmail address, update the cc list. > > On 8/10/26 5:12 PM, Dave Young wrote: >> Hi Kazu, >> >> While testing crash with recent mainline kernel, I applied your two patches. >> >> It does emit the errors mentioned, but crash failed to start, it failed in the task init path, I suspect below log with "crash -d 8": >> read_diskdump: paddr/pfn: 7b39d4c0/7b39d -> physical page is cached: 7b39d000 >> <read_diskdump: addr: ffff88807b39e000 paddr: 7b39e000 cnt: 192> >> read_diskdump: paddr/pfn: 7b39e000/7b39e -> physical page is cached: 7b39e000 >> task 1 0x8df932 >> please wait... (gathering task table data)<readmem: ffffffff81c25948, KVADDR, "xarray xa_head", 8, (FOE), 7ffcd7a168d0> >> <read_diskdump: addr: ffffffff81c25948 paddr: 2225948 cnt: 8> >> read_diskdump: paddr/pfn: 2225948/2225 -> cache physical page: 2225000 >> <readmem: ffff888003241b60, KVADDR, "xa_node shift", 1, (FOE), 7ffcd7a168cf> >> <read_diskdump: addr: ffff888003241b60 paddr: 3241b60 cnt: 1> >> read_diskdump: PAGE_EXCLUDED: paddr/pfn: 3241b60/3241 >> >> crash: page excluded: kernel virtual address: ffff888003241b60 type: "xa_node shift" >> >> >> And then I tried 6.19.0 kernel, crash works fine. Bisect finished with below result: >> bash-5.3$ git bisect bad >> Bisecting: 1 revision left to test after this (roughly 1 step) >> [d50569612c29215c5d1c64f47a65604ed265d2e9] mm: rename the 'compound_head' field in the 'struct page' to 'compound_info' >> bash-5.3$ git bisect view --oneline >> 8c846c879e22 mm: rework compound_head() for power-of-2 sizeof(struct page) >> 2969b42c8f99 LoongArch/mm: align vmemmap to maximal folio size >> 476849b0fba4 riscv/mm: align vmemmap to maximal folio size >> 67c79a5af051 mm: move set/clear_compound_head() next to compound_head() >> d50569612c29 (HEAD) mm: rename the 'compound_head' field in the 'struct page' to 'compound_info' >> bash-5.3$ git bisect skip >> There are only 'skip'ped commits left to test. >> The first bad commit could be any of: >> 476849b0fba4450f5adf22196bcff9c24c673bc4 >> 67c79a5af051f57339ecf383d3f67e200741ce20 >> 2969b42c8f994330fc020ec0d235aeb43bce317c >> d50569612c29215c5d1c64f47a65604ed265d2e9 >> 8c846c879e226c312c2c7a7bc1e323779903530f >> We cannot bisect more! >> >> >> Bisect skipped a few commits due to kernel can not compile, but the issue seems related to those commits. Any thoughts? I suspect your fixes are not enough? hmm, based on the kernel patches and the observed behavior, I suspect this makedumpfile patch, which fixes a wrong exclusion on Linux 7.1. https://github.com/makedumpfile/makedumpfile/commit/f50c015f150f93dc69f81... Does your makedumpfile include this patch?