On 11/08/2019 02:29 AM, Damien Le Moal wrote:
> Prevent recursion into the IO path under low memory conditions by using
> GFP_NOIO in place of GFP_KERNEL when allocating a new command with
> tcmu_alloc_cmd() and user ring space with tcmu_get_empty_block().
>
> Reported-by: Masato Suzuki <masato.suzuki@xxxxxxx>
> Signed-off-by: Damien Le Moal <damien.lemoal@xxxxxxx>
> ---
>
> Changes from v1:
> * Added reported-by tag
>
> drivers/target/target_core_user.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c
> index 35be1be87d2a..0b9dfa6b17bc 100644
> --- a/drivers/target/target_core_user.c
> +++ b/drivers/target/target_core_user.c
> @@ -499,7 +499,7 @@ static inline bool tcmu_get_empty_block(struct tcmu_dev *udev,
> schedule_delayed_work(&tcmu_unmap_work, 0);
>
> /* try to get new page from the mm */
> - page = alloc_page(GFP_KERNEL);
> + page = alloc_page(GFP_NOIO);
> if (!page)
> goto err_alloc;
>
> @@ -573,7 +573,7 @@ static struct tcmu_cmd *tcmu_alloc_cmd(struct se_cmd *se_cmd)
> struct tcmu_dev *udev = TCMU_DEV(se_dev);
> struct tcmu_cmd *tcmu_cmd;
>
> - tcmu_cmd = kmem_cache_zalloc(tcmu_cmd_cache, GFP_KERNEL);
> + tcmu_cmd = kmem_cache_zalloc(tcmu_cmd_cache, GFP_NOIO);
> if (!tcmu_cmd)
> return NULL;
>
> @@ -584,7 +584,7 @@ static struct tcmu_cmd *tcmu_alloc_cmd(struct se_cmd *se_cmd)
> tcmu_cmd_reset_dbi_cur(tcmu_cmd);
> tcmu_cmd->dbi_cnt = tcmu_cmd_get_block_cnt(tcmu_cmd);
> tcmu_cmd->dbi = kcalloc(tcmu_cmd->dbi_cnt, sizeof(uint32_t),
> - GFP_KERNEL);
> + GFP_NOIO);
> if (!tcmu_cmd->dbi) {
> kmem_cache_free(tcmu_cmd_cache, tcmu_cmd);
> return NULL;
>
Acked-by: Mike Christie <mchristi@xxxxxxxxxx>
We should also change tcmu_setup_cmd_timer so the gfp use in that code
path is consistent. I think we can do that in a separate patch later as
this one just fixes a specific bug.
