Re: [PATCH] habanalabs: fix double unlock on error in map_device_va()

From: Oded Gabbay
Date: Sun Jun 05 2022 - 06:06:46 EST


On Wed, May 25, 2022 at 3:25 PM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
>
> If hl_mmu_prefetch_cache_range() fails then this code calls
> mutex_unlock(&ctx->mmu_lock) when it's no longer holding the mutex.
>
> Fixes: 9e495e24003e ("habanalabs: do MMU prefetch as deferred work")
> Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> ---
> drivers/misc/habanalabs/common/memory.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/misc/habanalabs/common/memory.c b/drivers/misc/habanalabs/common/memory.c
> index 663dd7e589d4..d5e6500f8a1f 100644
> --- a/drivers/misc/habanalabs/common/memory.c
> +++ b/drivers/misc/habanalabs/common/memory.c
> @@ -1245,16 +1245,16 @@ static int map_device_va(struct hl_ctx *ctx, struct hl_mem_in *args, u64 *device
> rc = map_phys_pg_pack(ctx, ret_vaddr, phys_pg_pack);
> if (rc) {
> dev_err(hdev->dev, "mapping page pack failed for handle %u\n", handle);
> + mutex_unlock(&ctx->mmu_lock);
> goto map_err;
> }
>
> rc = hl_mmu_invalidate_cache_range(hdev, false, *vm_type | MMU_OP_SKIP_LOW_CACHE_INV,
> ctx->asid, ret_vaddr, phys_pg_pack->total_size);
> + mutex_unlock(&ctx->mmu_lock);
> if (rc)
> goto map_err;
>
> - mutex_unlock(&ctx->mmu_lock);
> -
> /*
> * prefetch is done upon user's request. it is performed in WQ as and so can
> * be outside the MMU lock. the operation itself is already protected by the mmu lock
> @@ -1283,8 +1283,6 @@ static int map_device_va(struct hl_ctx *ctx, struct hl_mem_in *args, u64 *device
> return rc;
>
> map_err:
> - mutex_unlock(&ctx->mmu_lock);
> -
> if (add_va_block(hdev, va_range, ret_vaddr,
> ret_vaddr + phys_pg_pack->total_size - 1))
> dev_warn(hdev->dev,
> --
> 2.35.1
>

Reviewed-by: Oded Gabbay <ogabbay@xxxxxxxxxx>
Applied to -next.
Thanks,
Oded