[Crash-utility] [PATCH] ARM: fix vtop -C for kernel addresses
Dave Anderson
anderson at redhat.com
Mon Mar 12 15:27:48 UTC 2012
----- Original Message -----
> On ARM, [vtop -C task] doesn't work correctly for kernel addresses: it
> always uses swapper_pg_dir when it should be using the task's
> mm->pgd.
>
> The patch below fixes this, and also makes ARM handle kernel threads
> like the other arches.
>
> Rabin
The patch makes "vtop -c" conform with the other architectures, so unless
Mika has a problem with it, consider it queued for crash-6.0.5.
Thanks,
Dave
> diff --git a/arm.c b/arm.c
> index 4be224e..cac9a9a 100644
> --- a/arm.c
> +++ b/arm.c
> @@ -1020,22 +1020,37 @@ static int
> arm_uvtop(struct task_context *tc, ulong uvaddr, physaddr_t *paddr,
> int verbose)
> {
> ulong *pgd;
> - ulong mm;
>
> if (!tc)
> error(FATAL, "current context invalid\n");
>
> *paddr = 0;
>
> - if (IS_KVADDR(uvaddr))
> - return arm_kvtop(tc, uvaddr, paddr, verbose);
> + if (is_kernel_thread(tc->task) && IS_KVADDR(uvaddr)) {
> + ulong active_mm;
>
> - mm = task_mm(tc->task, TRUE);
> - if (mm)
> - pgd = ULONG_PTR(tt->mm_struct + OFFSET(mm_struct_pgd));
> - else
> - readmem(tc->mm_struct + OFFSET(mm_struct_pgd), KVADDR, &pgd,
> - sizeof(long), "mm_struct pgd", FAULT_ON_ERROR);
> + readmem(tc->task + OFFSET(task_struct_active_mm),
> + KVADDR, &active_mm, sizeof(void *),
> + "task active_mm contents", FAULT_ON_ERROR);
> +
> + if (!active_mm)
> + error(FATAL,
> + "no active_mm for this kernel thread\n");
> +
> + readmem(active_mm + OFFSET(mm_struct_pgd),
> + KVADDR, &pgd, sizeof(long),
> + "mm_struct pgd", FAULT_ON_ERROR);
> + } else {
> + ulong mm;
> +
> + mm = task_mm(tc->task, TRUE);
> + if (mm)
> + pgd = ULONG_PTR(tt->mm_struct + OFFSET(mm_struct_pgd));
> + else
> + readmem(tc->mm_struct + OFFSET(mm_struct_pgd),
> + KVADDR, &pgd, sizeof(long), "mm_struct pgd",
> + FAULT_ON_ERROR);
> + }
>
> return arm_vtop(uvaddr, pgd, paddr, verbose);
> }
>
> --
> Crash-utility mailing list
> Crash-utility at redhat.com
> https://www.redhat.com/mailman/listinfo/crash-utility
>
More information about the Crash-utility
mailing list