2024-04-19 10:17:40

by Hailong Liu

Subject: [RFC PATCH v2] mm/vmalloc: fix return value of vb_alloc if size is 0.

From: "Hailong.Liu" <[email protected]>

vm_map_ram check return value of vb_alloc by IS_ERR. if
vm_map_ram(page, 0, 0) , vb_alloc(0, GFP_KERNEL) would return NULL
which cause kernel panic by vmap_pages_range_noflush。fix this by
return ERR_PTR(-EINVAL) if size is 0.

Signed-off-by: Hailong.Liu <[email protected]>
---
Changes since v1 [1]:
- Return ERR_PTR(-EINVAL) or not check IS_ERR_OR_NULL

BTW, Barry suggests me that if count is 0, return directly, in my
opinion, change return value is more resonable.

[1] https://lore.kernel.org/all/[email protected]/

mm/vmalloc.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index a3fedb3ee0db..c430a999805b 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2173,7 +2173,7 @@ static void *vb_alloc(unsigned long size, gfp_t gfp_mask)
* get_order(0) returns funny result. Just warn and terminate
* early.
*/
- return NULL;
+ return ERR_PTR(-EINVAL);
}
order = get_order(size);

--
2.34.1

2024-04-22 03:46:46

by Barry Song

Subject: Re: [RFC PATCH v2] mm/vmalloc: fix return value of vb_alloc if size is 0.

On Fri, Apr 19, 2024 at 6:17 PM <[email protected]> wrote:
>
> From: "Hailong.Liu" <[email protected]>
>
> vm_map_ram check return value of vb_alloc by IS_ERR. if
> vm_map_ram(page, 0, 0) , vb_alloc(0, GFP_KERNEL) would return NULL
> which cause kernel panic by vmap_pages_range_noflush。fix this by
> return ERR_PTR(-EINVAL) if size is 0.
>
> Signed-off-by: Hailong.Liu <[email protected]>
> ---
> Changes since v1 [1]:
> - Return ERR_PTR(-EINVAL) or not check IS_ERR_OR_NULL
>
> BTW, Barry suggests me that if count is 0, return directly, in my
> opinion, change return value is more resonable.
>
> [1] https://lore.kernel.org/all/84d7cd03-1cf8-401a-8edf-2524db0bd6d5@oppocom/
>
> mm/vmalloc.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index a3fedb3ee0db..c430a999805b 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2173,7 +2173,7 @@ static void *vb_alloc(unsigned long size, gfp_t gfp_mask)
> * get_order(0) returns funny result. Just warn and terminate
> * early.
> */
> - return NULL;
> + return ERR_PTR(-EINVAL);

might be ZERO_SIZE_PTR.

> }
> order = get_order(size);
>
> --
> 2.34.1

2024-04-22 09:41:18

by Uladzislau Rezki (Sony)

Subject: Re: [RFC PATCH v2] mm/vmalloc: fix return value of vb_alloc if size is 0.

On Fri, Apr 19, 2024 at 06:16:43PM +0800, [email protected] wrote:
> From: "Hailong.Liu" <[email protected]>
>
> vm_map_ram check return value of vb_alloc by IS_ERR. if
> vm_map_ram(page, 0, 0) , vb_alloc(0, GFP_KERNEL) would return NULL
> which cause kernel panic by vmap_pages_range_noflush。fix this by
> return ERR_PTR(-EINVAL) if size is 0.
>
> Signed-off-by: Hailong.Liu <[email protected]>
> ---
> Changes since v1 [1]:
> - Return ERR_PTR(-EINVAL) or not check IS_ERR_OR_NULL
>
> BTW, Barry suggests me that if count is 0, return directly, in my
> opinion, change return value is more resonable.
>
> [1] https://lore.kernel.org/all/[email protected]/
>
> mm/vmalloc.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index a3fedb3ee0db..c430a999805b 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2173,7 +2173,7 @@ static void *vb_alloc(unsigned long size, gfp_t gfp_mask)
> * get_order(0) returns funny result. Just warn and terminate
> * early.
> */
> - return NULL;
> + return ERR_PTR(-EINVAL);
> }
> order = get_order(size);
>
> --
> 2.34.1
>
Reviewed-by: Uladzislau Rezki (Sony) <[email protected]>

--
Uladzislau Rezki

2024-04-22 10:43:48

by Hailong Liu

Subject: Re: [RFC PATCH v2] mm/vmalloc: fix return value of vb_alloc if size is 0.

On Mon, 22. Apr 11:46, Barry Song wrote:
> On Fri, Apr 19, 2024 at 6:17 PM <[email protected]> wrote:
> >
> > From: "Hailong.Liu" <[email protected]>
> >
> > vm_map_ram check return value of vb_alloc by IS_ERR. if
> > vm_map_ram(page, 0, 0) , vb_alloc(0, GFP_KERNEL) would return NULL
> > which cause kernel panic by vmap_pages_range_noflush。fix this by
> > return ERR_PTR(-EINVAL) if size is 0.
> >
> > Signed-off-by: Hailong.Liu <[email protected]>
> > ---
> > Changes since v1 [1]:
> > - Return ERR_PTR(-EINVAL) or not check IS_ERR_OR_NULL
> >
> > BTW, Barry suggests me that if count is 0, return directly, in my
> > opinion, change return value is more resonable.
> >
> > [1] https://lore.kernel.org/all/[email protected]/
> >
> > mm/vmalloc.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > index a3fedb3ee0db..c430a999805b 100644
> > --- a/mm/vmalloc.c
> > +++ b/mm/vmalloc.c
> > @@ -2173,7 +2173,7 @@ static void *vb_alloc(unsigned long size, gfp_t gfp_mask)
> > * get_order(0) returns funny result. Just warn and terminate
> > * early.
> > */
> > - return NULL;
> > + return ERR_PTR(-EINVAL);
>
> might be ZERO_SIZE_PTR.

Hi Barry,
Hi Barry, I use ERR_PTR(-EINVAL) to keep consistency with the return
value of the function alloc_vmap_area.

if (unlikely(!size || offset_in_page(size) || !is_power_of_2(align)))
return ERR_PTR(-EINVAL);

IMO, ZERO_SIZE_PTR is used by slab and can not be catched by IS_ERR().

--
Best Regards,
Hailong.

2024-04-23 01:10:16

by Barry Song

Subject: Re: [RFC PATCH v2] mm/vmalloc: fix return value of vb_alloc if size is 0.

On Mon, Apr 22, 2024 at 6:39 PM Hailong Liu <[email protected]> wrote:
>
> On Mon, 22. Apr 11:46, Barry Song wrote:
> > On Fri, Apr 19, 2024 at 6:17 PM <[email protected]> wrote:
> > >
> > > From: "Hailong.Liu" <[email protected]>
> > >
> > > vm_map_ram check return value of vb_alloc by IS_ERR. if
> > > vm_map_ram(page, 0, 0) , vb_alloc(0, GFP_KERNEL) would return NULL
> > > which cause kernel panic by vmap_pages_range_noflush。fix this by
> > > return ERR_PTR(-EINVAL) if size is 0.
> > >
> > > Signed-off-by: Hailong.Liu <[email protected]>
> > > ---
> > > Changes since v1 [1]:
> > > - Return ERR_PTR(-EINVAL) or not check IS_ERR_OR_NULL
> > >
> > > BTW, Barry suggests me that if count is 0, return directly, in my
> > > opinion, change return value is more resonable.
> > >
> > > [1] https://lore.kernel.org/all/[email protected]/
> > >
> > > mm/vmalloc.c | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > > index a3fedb3ee0db..c430a999805b 100644
> > > --- a/mm/vmalloc.c
> > > +++ b/mm/vmalloc.c
> > > @@ -2173,7 +2173,7 @@ static void *vb_alloc(unsigned long size, gfp_t gfp_mask)
> > > * get_order(0) returns funny result. Just warn and terminate
> > > * early.
> > > */
> > > - return NULL;
> > > + return ERR_PTR(-EINVAL);
> >
> > might be ZERO_SIZE_PTR.
>
> Hi Barry,
> Hi Barry, I use ERR_PTR(-EINVAL) to keep consistency with the return
> value of the function alloc_vmap_area.
>
> if (unlikely(!size || offset_in_page(size) || !is_power_of_2(align)))
> return ERR_PTR(-EINVAL);
>
> IMO, ZERO_SIZE_PTR is used by slab and can not be catched by IS_ERR().

Ok. it seems not worth to change two places.

Reviewed-by: Barry Song <[email protected]>

>
> --
> Best Regards,
> Hailong.