sg_alloc_table_chained() is called by several drivers, but if it is
called before sg_pool_init(), it results in a NULL pointer dereference
in sg_pool_alloc().
Since commit 9b1d6c895002 ("lib: scatterlist: move SG pool code from
SCSI driver to lib/sg_pool.c"), we rely on module_init(sg_pool_init)
is invoked before other module_init calls but this assumption is
fragile.
I slightly changed the link order while Kbuild refactoring Kbuild,
then uncovered this issue. I should keep the current link order, but
depending on a specific call order among module_init is so fragine.
We usually define the init order by specifying *_initcall correctly,
or delay the driver probing by returning -EPROBE_DEFER.
Change module_initcall() to subsys_initcall(), and also delete the
pointless module_exit() because lib/sg_pool.c is always compiled as
built-in. (CONFIG_SG_POOL is bool)
Link: https://lore.kernel.org/all/[email protected]/
Link: https://lore.kernel.org/all/[email protected]/
Fixes: 9b1d6c895002 ("lib: scatterlist: move SG pool code from SCSI driver to lib/sg_pool.c")
Reported-by: Guenter Roeck <[email protected]>
Reported-by: Marek Szyprowski <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>
---
I am sending this to DMA subsystem because I did not find
a corresponding one for lib/sg_pool.c
lib/sg_pool.c | 16 ++--------------
1 file changed, 2 insertions(+), 14 deletions(-)
diff --git a/lib/sg_pool.c b/lib/sg_pool.c
index a0b1a52cd6f7..9bfe60ca3f37 100644
--- a/lib/sg_pool.c
+++ b/lib/sg_pool.c
@@ -1,5 +1,5 @@
// SPDX-License-Identifier: GPL-2.0-only
-#include <linux/module.h>
+#include <linux/init.h>
#include <linux/scatterlist.h>
#include <linux/mempool.h>
#include <linux/slab.h>
@@ -177,16 +177,4 @@ static __init int sg_pool_init(void)
return -ENOMEM;
}
-static __exit void sg_pool_exit(void)
-{
- int i;
-
- for (i = 0; i < SG_MEMPOOL_NR; i++) {
- struct sg_pool *sgp = sg_pools + i;
- mempool_destroy(sgp->pool);
- kmem_cache_destroy(sgp->slab);
- }
-}
-
-module_init(sg_pool_init);
-module_exit(sg_pool_exit);
+subsys_initcall(sg_pool_init);
--
2.34.1
On 23.09.2022 13:38, Masahiro Yamada wrote:
> sg_alloc_table_chained() is called by several drivers, but if it is
> called before sg_pool_init(), it results in a NULL pointer dereference
> in sg_pool_alloc().
>
> Since commit 9b1d6c895002 ("lib: scatterlist: move SG pool code from
> SCSI driver to lib/sg_pool.c"), we rely on module_init(sg_pool_init)
> is invoked before other module_init calls but this assumption is
> fragile.
>
> I slightly changed the link order while Kbuild refactoring Kbuild,
> then uncovered this issue. I should keep the current link order, but
> depending on a specific call order among module_init is so fragine.
>
> We usually define the init order by specifying *_initcall correctly,
> or delay the driver probing by returning -EPROBE_DEFER.
>
> Change module_initcall() to subsys_initcall(), and also delete the
> pointless module_exit() because lib/sg_pool.c is always compiled as
> built-in. (CONFIG_SG_POOL is bool)
>
> Link: https://lore.kernel.org/all/[email protected]/
> Link: https://lore.kernel.org/all/[email protected]/
> Fixes: 9b1d6c895002 ("lib: scatterlist: move SG pool code from SCSI driver to lib/sg_pool.c")
> Reported-by: Guenter Roeck <[email protected]>
> Reported-by: Marek Szyprowski <[email protected]>
> Signed-off-by: Masahiro Yamada <[email protected]>
> ---
>
> I am sending this to DMA subsystem because I did not find
> a corresponding one for lib/sg_pool.c
Feel free to add:
Tested-by: Marek Szyprowski <[email protected]>
> lib/sg_pool.c | 16 ++--------------
> 1 file changed, 2 insertions(+), 14 deletions(-)
>
> diff --git a/lib/sg_pool.c b/lib/sg_pool.c
> index a0b1a52cd6f7..9bfe60ca3f37 100644
> --- a/lib/sg_pool.c
> +++ b/lib/sg_pool.c
> @@ -1,5 +1,5 @@
> // SPDX-License-Identifier: GPL-2.0-only
> -#include <linux/module.h>
> +#include <linux/init.h>
> #include <linux/scatterlist.h>
> #include <linux/mempool.h>
> #include <linux/slab.h>
> @@ -177,16 +177,4 @@ static __init int sg_pool_init(void)
> return -ENOMEM;
> }
>
> -static __exit void sg_pool_exit(void)
> -{
> - int i;
> -
> - for (i = 0; i < SG_MEMPOOL_NR; i++) {
> - struct sg_pool *sgp = sg_pools + i;
> - mempool_destroy(sgp->pool);
> - kmem_cache_destroy(sgp->slab);
> - }
> -}
> -
> -module_init(sg_pool_init);
> -module_exit(sg_pool_exit);
> +subsys_initcall(sg_pool_init);
Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland
On 2022-09-23 12:38, Masahiro Yamada wrote:
> sg_alloc_table_chained() is called by several drivers, but if it is
> called before sg_pool_init(), it results in a NULL pointer dereference
> in sg_pool_alloc().
>
> Since commit 9b1d6c895002 ("lib: scatterlist: move SG pool code from
> SCSI driver to lib/sg_pool.c"), we rely on module_init(sg_pool_init)
> is invoked before other module_init calls but this assumption is
> fragile.
>
> I slightly changed the link order while Kbuild refactoring Kbuild,
> then uncovered this issue. I should keep the current link order, but
> depending on a specific call order among module_init is so fragine.
>
> We usually define the init order by specifying *_initcall correctly,
> or delay the driver probing by returning -EPROBE_DEFER.
>
> Change module_initcall() to subsys_initcall(), and also delete the
> pointless module_exit() because lib/sg_pool.c is always compiled as
> built-in. (CONFIG_SG_POOL is bool)
Makes sense to me. Short of having some cool-but-impractically-complex
system to derive a dependency graph from the config and compute an
initcall order from that, initialising helper library code at an earlier
step than drivers certainly seems like the next-best option, and subsys
doesn't seem inappropriate for the nature of this code. FWIW,
Reviewed-by: Robin Murphy <[email protected]>
> Link: https://lore.kernel.org/all/[email protected]/
> Link: https://lore.kernel.org/all/[email protected]/
> Fixes: 9b1d6c895002 ("lib: scatterlist: move SG pool code from SCSI driver to lib/sg_pool.c")
> Reported-by: Guenter Roeck <[email protected]>
> Reported-by: Marek Szyprowski <[email protected]>
> Signed-off-by: Masahiro Yamada <[email protected]>
> ---
>
> I am sending this to DMA subsystem because I did not find
> a corresponding one for lib/sg_pool.c
>
> lib/sg_pool.c | 16 ++--------------
> 1 file changed, 2 insertions(+), 14 deletions(-)
>
> diff --git a/lib/sg_pool.c b/lib/sg_pool.c
> index a0b1a52cd6f7..9bfe60ca3f37 100644
> --- a/lib/sg_pool.c
> +++ b/lib/sg_pool.c
> @@ -1,5 +1,5 @@
> // SPDX-License-Identifier: GPL-2.0-only
> -#include <linux/module.h>
> +#include <linux/init.h>
> #include <linux/scatterlist.h>
> #include <linux/mempool.h>
> #include <linux/slab.h>
> @@ -177,16 +177,4 @@ static __init int sg_pool_init(void)
> return -ENOMEM;
> }
>
> -static __exit void sg_pool_exit(void)
> -{
> - int i;
> -
> - for (i = 0; i < SG_MEMPOOL_NR; i++) {
> - struct sg_pool *sgp = sg_pools + i;
> - mempool_destroy(sgp->pool);
> - kmem_cache_destroy(sgp->slab);
> - }
> -}
> -
> -module_init(sg_pool_init);
> -module_exit(sg_pool_exit);
> +subsys_initcall(sg_pool_init);
Thanks, applied with the typo fixes included.
On Fri, Sep 23, 2022 at 08:38:35PM +0900, Masahiro Yamada wrote:
> sg_alloc_table_chained() is called by several drivers, but if it is
> called before sg_pool_init(), it results in a NULL pointer dereference
> in sg_pool_alloc().
>
> Since commit 9b1d6c895002 ("lib: scatterlist: move SG pool code from
> SCSI driver to lib/sg_pool.c"), we rely on module_init(sg_pool_init)
> is invoked before other module_init calls but this assumption is
> fragile.
>
> I slightly changed the link order while Kbuild refactoring Kbuild,
> then uncovered this issue. I should keep the current link order, but
> depending on a specific call order among module_init is so fragine.
>
> We usually define the init order by specifying *_initcall correctly,
> or delay the driver probing by returning -EPROBE_DEFER.
>
> Change module_initcall() to subsys_initcall(), and also delete the
> pointless module_exit() because lib/sg_pool.c is always compiled as
> built-in. (CONFIG_SG_POOL is bool)
>
> Link: https://lore.kernel.org/all/[email protected]/
> Link: https://lore.kernel.org/all/[email protected]/
> Fixes: 9b1d6c895002 ("lib: scatterlist: move SG pool code from SCSI driver to lib/sg_pool.c")
> Reported-by: Guenter Roeck <[email protected]>
> Reported-by: Marek Szyprowski <[email protected]>
> Signed-off-by: Masahiro Yamada <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
Guenter