On Wed, Nov 04, 2026 at 08:25:32PM +0100, Christian Brauner wrote:
> A capture kernel reads the memory of the crashed kernel, and that memory
> may well have uncorrectable errors. On powerpc a machine check in the
> kernel is only recovered from if the instruction that took it has an
> exception table entry (cf. mce_common_process_ue()). copy_oldmem_page()
> copies with copy_to_iter(). For a read() of /proc/vmcore that is a user
> copy, all of whose loads have exception table entries, so the machine
> check turns into a short copy and read_from_oldmem() returns -EFAULT.
>
> But since commit 4a22fd20379c ("vmcore: convert __read_vmcore to use an
> iov_iter") /proc/vmcore can be spliced too, and splice(), sendfile()
> and copy_file_range() copy into kernel pages with memcpy(), which has
> no exception table entries. A machine check there panics the capture
> kernel. The same goes for the reads of the ELF headers and notes that
> vmcore does itself, and those have been memcpy() for as long as
> copy_oldmem_page() has had its kernel buffer branch, since commit
> 54c32021eb6f ("[PATCH] powerpc: Add arch-dependent copy_oldmem_page").
>
> Copy with _copy_mc_to_iter() instead. PPC64 selects ARCH_HAS_COPY_MC
> and there it uses copy_mc_generic() for kernel and user destinations
> alike, which has the exception table entries. Everywhere else it is
> _copy_to_iter().
>
> Reported-by: Chris Mason <[email protected]>
> Fixes: 54c32021eb6f ("[PATCH] powerpc: Add arch-dependent copy_oldmem_page")
> Fixes: 4a22fd20379c ("vmcore: convert __read_vmcore to use an iov_iter")
> Cc: [email protected]
> Signed-off-by: Christian Brauner (Amutable) <[email protected]>
Reviewed-by: Breno Leitao <[email protected]>
The change is the right direction, and x86_64 has the same gap.
arch/x86/kernel/crash_dump_64.c:__copy_oldmem_page() copies with
copy_to_iter(), the same as the powerpc code this patch changes. x86_64
selects ARCH_HAS_COPY_MC, and _copy_mc_to_iter() is already available,
with the _copy_to_iter() fallback in include/linux/uio.h for arches
without it. The plain and encrypted paths share __copy_oldmem_page(), so
one conversion there covers both.
PS: I found this because I am looking at x856 production fleet kdump
success rate , where first kernel died of an MCE very often produce no
usable vmcore at all, and makedumpfile never finishes due to this type
of problem.
