[Crash-utility] Re: [PATCH v3 3/4] xarray: add large folio support
Right, sorry for that. I mistakenly used an old version of your patch
for the regression test, which led to the false result. I will redo
the regression test and check the results, please wait a while. Sorry
again for that.
Thanks,
Tao Liu
On Wed, Mar 4, 2026 at 3:49 PM Huang Shijie <huangsj(a)hygon.cn> wrote:
On Wed, Mar 04, 2026 at 02:45:10PM +1300, Tao Liu wrote:
> Hi Huang,
>
> I still encounter some regressions for v3:
>
> crash: invalid structure member offset: folio__flags_1
> FILE: memory.c LINE: 20445 FUNCTION: folio_order()
>
> [/root/crash-dev/crash] error trace: 9be19e => a12519 => a05a6f => adf8fd
> Detaching from process 3369535
>
> adf8fd: OFFSET_verify+164
> Detaching from process 3369536
> a05a6f: folio_order+156
> Detaching from process 3369537
> a12519: do_xarray_count+48
> Detaching from process 3369538
> 9be19e: do_xarray_iter+334
>
>
> Child exited with status 1
>
> Thread 1 "crash" hit Breakpoint 1, folio_order
> (folio=18446626868456917760) at memory.c:20445
> 20445 readmem(folio + OFFSET(folio__flags_1), KVADDR, &v, sizeof(ulong),
By the way, your vmlinux/vmcore is for 4.18.
In the v3 code for folio_order:
int
folio_order(ulong folio)
{
ulong v = 0;
int PG_head = 16;
if (THIS_KERNEL_VERSION < LINUX(5,17,0))
return 0; <---- return here.
The code should not run to the line:
readmem(folio + OFFSET(folio__flags_1), KVADDR, &v, sizeof(ulong),
Thanks
Huang Shijie