On 29. 08. 19, 14:13, François Valenduc wrote:

The following patch causes a build failure:

Author: Henry Burns henryburns@google.com Date: Sat Aug 24 17:55:06 2019 -0700

mm/zsmalloc.c: fix race condition in zs_destroy_pool

So this is f6d997de0883 in 5.2.11 and 701d678599d0 in upstream.

I get this error:

CALL scripts/checksyscalls.sh CALL scripts/atomic/check-atomics.sh DESCEND objtool CHK include/generated/compile.h CC mm/zsmalloc.o In file included from ./include/linux/mmzone.h:10:0, from ./include/linux/gfp.h:6, from ./include/linux/umh.h:4, from ./include/linux/kmod.h:9, from ./include/linux/module.h:13, from mm/zsmalloc.c:33: mm/zsmalloc.c: In function ‘zs_create_pool’: mm/zsmalloc.c:2435:27: error: ‘struct zs_pool’ has no member named ‘migration_wait’ init_waitqueue_head(&pool->migration_wait);

Obviously, as this is not protected by #ifdef CONFIG_COMPACTION ... #endif as is its definition in the structure (and its other uses).

./include/linux/wait.h:67:26: note: in definition of macro ‘init_waitqueue_head’ __init_waitqueue_head((wq_head), #wq_head, &__key); \ ^~~~~~~ scripts/Makefile.build:278: recipe for target 'mm/zsmalloc.o' failed make[1]: *** [mm/zsmalloc.o] Error 1 Makefile:1073: recipe for target 'mm' failed

You can find my configuration file attached.

You forgot to attach it, but you have CONFIG_COMPACTION=n, I assume.

Does anybody have any idea about this ?

Sure, this will fix it (or turn on compaction): --- a/mm/zsmalloc.c +++ b/mm/zsmalloc.c @@ -2413,7 +2413,9 @@ struct zs_pool *zs_create_pool(const char *name) if (!pool->name) goto err;

+#ifdef CONFIG_COMPACTION init_waitqueue_head(&pool->migration_wait); +#endif

if (create_cache(pool)) goto err;

thanks,

On (08/29/19 14:28), Jiri Slaby wrote: [..]

as is its definition in the structure (and its other uses).

...

./include/linux/wait.h:67:26: note: in definition of macro ‘init_waitqueue_head’ __init_waitqueue_head((wq_head), #wq_head, &__key); \ ^~~~~~~ scripts/Makefile.build:278: recipe for target 'mm/zsmalloc.o' failed make[1]: *** [mm/zsmalloc.o] Error 1 Makefile:1073: recipe for target 'mm' failed

You can find my configuration file attached.

You forgot to attach it, but you have CONFIG_COMPACTION=n, I assume.

...

Does anybody have any idea about this ?

Sure, this will fix it (or turn on compaction): --- a/mm/zsmalloc.c +++ b/mm/zsmalloc.c @@ -2413,7 +2413,9 @@ struct zs_pool *zs_create_pool(const char *name) if (!pool->name) goto err;

+#ifdef CONFIG_COMPACTION init_waitqueue_head(&pool->migration_wait); +#endif

The fix is correct. I believe Andrew already has the same patch in his tree.

-ss

On Mon, Sep 02, 2019 at 08:01:45PM +0200, François Valenduc wrote:

Le 29/08/19 à 14:32, Sergey Senozhatsky a écrit :

...

On (08/29/19 14:28), Jiri Slaby wrote: [..]

...

as is its definition in the structure (and its other uses).

...

./include/linux/wait.h:67:26: note: in definition of macro ‘init_waitqueue_head’ __init_waitqueue_head((wq_head), #wq_head, &__key); \ ^~~~~~~ scripts/Makefile.build:278: recipe for target 'mm/zsmalloc.o' failed make[1]: *** [mm/zsmalloc.o] Error 1 Makefile:1073: recipe for target 'mm' failed

You can find my configuration file attached.

You forgot to attach it, but you have CONFIG_COMPACTION=n, I assume.

...

Does anybody have any idea about this ?

Sure, this will fix it (or turn on compaction): --- a/mm/zsmalloc.c +++ b/mm/zsmalloc.c @@ -2413,7 +2413,9 @@ struct zs_pool *zs_create_pool(const char *name) if (!pool->name) goto err;

+#ifdef CONFIG_COMPACTION init_waitqueue_head(&pool->migration_wait); +#endif

The fix is correct. I believe Andrew already has the same patch in his tree.

-ss

No reaction of the stable team on this ? Meanwhile, the fix is in the mainline tree (commit 441e254cd40dc03beec3c650ce6ce6074bc6517f).

Give us a chance to catch up! :)

Hopefully it can be included in the next stable releases. However, if I read the config text, memory compaction should always be enabled, even if it is disabled by default.

It's now queued up everywhere, thanks.

greg k-h