Re: client side "current->alloc_tag not set WARNING: ./include/linux/alloc_tag.h:161" at accessing CIFS share

[Hao Ge <hao.ge@linux.dev>]

Hi Paulo


Sorry to jump in.

On 2026/9/2 07:19, Paulo Alcantara wrote:
> Erhard Furtner <erhard_f@mailbox.org> writes:
> 
>> Getting this "current->alloc_tag not set WARNING:" every time on my 
>> client PC when I open files on the CIFS share (exported via KSMBD) from 
>> my host PC:
> 
> The warning is triggered by running the following in
> netfs_alloc_request() with CONFIG_MEM_ALLOC_PROFILING_DEBUG=y kernels:
> 
>         rreq = mempool->alloc(gfp, mempool->pool_data);
> 
> for the non-writeback path.
> 
> @mempool->alloc is set to mempool_alloc_slab(), which calls
> kmem_cache_alloc_noprof() directly and expects the caller to have
> already set current->alloc_tag.
> 
> It doesn't occur in the writeback path, however, because we call
> mempool_alloc() macro that wraps to
> 
>         alloc_hooks(mempool_alloc_noprof(...))
> 
> and it sets current->alloc_tag before the allocation.
>

Right - and there's one more instance of the same pattern.
Erhard's second warning (alloc_tag was not set on free) comes from
netfs_folioq_alloc(), which does the same for the folio_queue:
	
	fq = netfs_folioq_pool.alloc(gfp, netfs_folioq_pool.pool_data);

When I first saw the report I wondered why no other mempool user trips over this,
so I had a look around the tree:

Everyone else gets their elements through the mempool_alloc() macro, so the tag is
set at their callsite and the callbacks see a valid current->alloc_tag.

I went back to the commit that introduced the raw call, 1d78d56c43ef ("netfs: Fix folio_queue ENOMEM in
writeback by adding a mempool") and OK, I think I roughly understand the story now.

Here is my proposed fix patch. I don't have a test environment handy though. What do you think?

diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h
index 420ee7b26580..373be970ec31 100644
--- a/fs/netfs/internal.h
+++ b/fs/netfs/internal.h
@@ -45,6 +45,9 @@ extern mempool_t netfs_request_pool;
 extern mempool_t netfs_subrequest_pool;
 extern mempool_t netfs_folioq_pool;
 
+#define mempool_alloc_element(_pool, _gfp)				\
+	alloc_hooks((_pool)->alloc(_gfp, (_pool)->pool_data))
+
 #ifdef CONFIG_PROC_FS
 static inline void netfs_proc_add_rreq(struct netfs_io_request *rreq)
 {
diff --git a/fs/netfs/objects.c b/fs/netfs/objects.c
index 01461a74642d..18f5c64d404e 100644
--- a/fs/netfs/objects.c
+++ b/fs/netfs/objects.c
@@ -34,7 +34,7 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
 
 		rreq = mempool_alloc(mempool, gfp);
 	} else {
-		rreq = mempool->alloc(gfp, mempool->pool_data);
+		rreq = mempool_alloc_element(mempool, gfp);
 		if (!rreq)
 			return ERR_PTR(-ENOMEM);
 	}
@@ -206,7 +206,7 @@ struct netfs_io_subrequest *netfs_alloc_subrequest(struct netfs_io_request *rreq
 	struct kmem_cache *cache = mempool->pool_data;
 
 	if (rreq->gfp == GFP_KERNEL)
-		subreq = mempool->alloc(rreq->gfp, mempool->pool_data);
+		subreq = mempool_alloc_element(mempool, rreq->gfp);
 	else
 		subreq = mempool_alloc(mempool, rreq->gfp);
 	if (!subreq)
diff --git a/fs/netfs/rolling_buffer.c b/fs/netfs/rolling_buffer.c
index 8c0026836f9c..d1ae6e081664 100644
--- a/fs/netfs/rolling_buffer.c
+++ b/fs/netfs/rolling_buffer.c
@@ -29,7 +29,7 @@ struct folio_queue *netfs_folioq_alloc(unsigned int rreq_id, gfp_t gfp,
 	struct folio_queue *fq;
 
 	if (gfp == GFP_KERNEL)
-		fq = netfs_folioq_pool.alloc(gfp, netfs_folioq_pool.pool_data);
+		fq = mempool_alloc_element(&netfs_folioq_pool, gfp);
 	else
 		fq = mempool_alloc(&netfs_folioq_pool, gfp);
 	if (fq) {
-- 
2.25.1

Thanks
Best Regards
Hao

> I'll look for a possible solution and then let you know.
> 
> Thanks for the report.