[Intel-xe] [PATCH] drm/xe: Prevent null ptr dereference

[Summers, Stuart]

On Sat, 2023-11-04 at 02:49 +0530, Pallavi Mishra wrote:
> In scenarios where vm is not initialized, use xe_migrate
> to get the address.
> 
> Signed-off-by: Pallavi Mishra <pallavi.mishra at intel.com>
> ---
>  drivers/gpu/drm/xe/xe_migrate.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_migrate.c
> b/drivers/gpu/drm/xe/xe_migrate.c
> index e6a31f90ebdb..f79ec8e4ae4e 100644
> --- a/drivers/gpu/drm/xe/xe_migrate.c
> +++ b/drivers/gpu/drm/xe/xe_migrate.c
> @@ -1296,7 +1296,10 @@ xe_migrate_update_pgtables(struct xe_migrate
> *m,
>  
>                         xe_tile_assert(tile, pt_bo->size == SZ_4K);
>  
> -                       addr = vm->pt_ops->pte_encode_bo(pt_bo, 0,
> pat_index, 0);
> +                       if (vm)
> +                               addr = vm->pt_ops-
> >pte_encode_bo(pt_bo, 0, pat_index, 0);
> +                       else
> +                               addr = m->q->vm->pt_ops-
> >pte_encode_bo(pt_bo, 0, pat_index, 0);
>                         bb->cs[bb->len++] = lower_32_bits(addr);
>                         bb->cs[bb->len++] = upper_32_bits(addr);
>                 }

I also see this in the routine in question:
        if (first_munmap_rebind) {
                err = job_add_deps(job, &vm->resv,
                                   DMA_RESV_USAGE_BOOKKEEP);
                if (err)
                        goto err_job;
        }

so looks like we have a potential for two NPDs here.

Also in xe_migrate_update_pgtables_cpu:
        if (wait_vm && !dma_resv_test_signaled(&vm->resv,
                                              
DMA_RESV_USAGE_BOOKKEEP))
                return ERR_PTR(-ETIME);

Thanks,
Stuart