[PATCH] btrfs: always pass __GFP_NOWARN from add_ra_bio_pages()

[PATCH] btrfs: always pass __GFP_NOWARN from add_ra_bio_pages()

[Calvin Owens]

A build workload newly prints order-0 allocation failures on 7.1-rc1:

    sh: page allocation failure: order:0
    mode:0x14084a(__GFP_HIGHMEM|__GFP_MOVABLE|__GFP_IO|__GFP_KSWAPD_RECLAIM|
                  __GFP_COMP|__GFP_HARDWALL)
    CPU: 27 UID: 1000 PID: 855540 Comm: sh Not tainted 7.1.0-rc1-llvm-00058-gdca922e019dd #1 PREEMPTLAZY
    Call Trace:
     <TASK>
     dump_stack_lvl+0x50/0x70
     warn_alloc+0xeb/0x100
     __alloc_pages_slowpath+0x567/0x5a0
     ? filemap_get_entry+0x11a/0x140
     __alloc_frozen_pages_noprof+0x249/0x2d0
     alloc_pages_mpol+0xe4/0x180
     folio_alloc_noprof+0x80/0xa0
     add_ra_bio_pages+0x13c/0x4b0
     btrfs_submit_compressed_read+0x229/0x300
     submit_one_bio+0x9e/0xe0
     btrfs_readahead+0x185/0x1a0
     [...]

    (lldb) source list -a add_ra_bio_pages+0x13c
    .../vmlinux.unstripped add_ra_bio_pages + 316 at .../fs/btrfs/compression.c:454:8
       451
       452                  folio = filemap_alloc_folio(mapping_gfp_constraint(mapping, constraint_gfp),
       453                                              0, NULL);
    -> 454                  if (!folio)
       455                          break;

I can reproduce this consistently by running a memory hog concurrently
with a buffered writer on a machine with a very large amount of swap.

Commit 7ae37b2c94ed ("btrfs: prevent direct reclaim during compressed
readahead") clearly intended to suppress these warnings. But because the
mask set in the address_space with mapping_set_gfp_mask() doesn't include
__GFP_NOWARN, mapping_gfp_constraint() removes it from constraint_gfp
before it is passed to filemap_alloc_folio().

Fix by refactoring the code to add __GFP_NOWARN after the call to
mapping_gfp_constraint().

Fixes: 7ae37b2c94ed ("btrfs: prevent direct reclaim during compressed readahead")
Signed-off-by: Calvin Owens <calvin@wbinvd.org>
---
 fs/btrfs/compression.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)

diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index b2393a48a8fe..d29a42ff8653 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -408,21 +408,19 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 	end_index = (i_size_read(inode) - 1) >> PAGE_SHIFT;
 
 	/*
-	 * Avoid direct reclaim when the caller does not allow it.  Since
-	 * add_ra_bio_pages() is always speculative, suppress allocation warnings
-	 * in either case.
+	 * Avoid direct reclaim when the caller does not allow it.
 	 */
+	constraint_gfp = ~__GFP_FS;
+	cache_gfp = GFP_NOFS | __GFP_NOWARN;
 	if (!direct_reclaim) {
-		constraint_gfp = ~(__GFP_FS | __GFP_DIRECT_RECLAIM) | __GFP_NOWARN;
-		cache_gfp = (GFP_NOFS & ~__GFP_DIRECT_RECLAIM) | __GFP_NOWARN;
-	} else {
-		constraint_gfp = (~__GFP_FS) | __GFP_NOWARN;
-		cache_gfp = GFP_NOFS | __GFP_NOWARN;
+		constraint_gfp &= ~__GFP_DIRECT_RECLAIM;
+		cache_gfp &= ~__GFP_DIRECT_RECLAIM;
 	}
 
 	while (cur < compressed_end) {
 		pgoff_t page_end;
 		pgoff_t pg_index = cur >> PAGE_SHIFT;
+		gfp_t masked_constraint_gfp;
 		u32 add_size;
 
 		if (pg_index > end_index)
@@ -449,8 +447,14 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 			continue;
 		}
 
-		folio = filemap_alloc_folio(mapping_gfp_constraint(mapping, constraint_gfp),
-					    0, NULL);
+		/*
+		 * Since add_ra_bio_pages() is always speculative, suppress
+		 * allocation warnings.
+		 */
+		masked_constraint_gfp = mapping_gfp_constraint(mapping, constraint_gfp);
+		masked_constraint_gfp |= __GFP_NOWARN;
+
+		folio = filemap_alloc_folio(masked_constraint_gfp, 0, NULL);
 		if (!folio)
 			break;
 
-- 
2.47.3

Re: [PATCH] btrfs: always pass __GFP_NOWARN from add_ra_bio_pages()

[David Sterba]

On Wed, Apr 29, 2026 at 12:10:25AM -0700, Calvin Owens wrote:
> A build workload newly prints order-0 allocation failures on 7.1-rc1:
> 
>     sh: page allocation failure: order:0
>     mode:0x14084a(__GFP_HIGHMEM|__GFP_MOVABLE|__GFP_IO|__GFP_KSWAPD_RECLAIM|
>                   __GFP_COMP|__GFP_HARDWALL)
>     CPU: 27 UID: 1000 PID: 855540 Comm: sh Not tainted 7.1.0-rc1-llvm-00058-gdca922e019dd #1 PREEMPTLAZY
>     Call Trace:
>      <TASK>
>      dump_stack_lvl+0x50/0x70
>      warn_alloc+0xeb/0x100
>      __alloc_pages_slowpath+0x567/0x5a0
>      ? filemap_get_entry+0x11a/0x140
>      __alloc_frozen_pages_noprof+0x249/0x2d0
>      alloc_pages_mpol+0xe4/0x180
>      folio_alloc_noprof+0x80/0xa0
>      add_ra_bio_pages+0x13c/0x4b0
>      btrfs_submit_compressed_read+0x229/0x300
>      submit_one_bio+0x9e/0xe0
>      btrfs_readahead+0x185/0x1a0
>      [...]
> 
>     (lldb) source list -a add_ra_bio_pages+0x13c
>     .../vmlinux.unstripped add_ra_bio_pages + 316 at .../fs/btrfs/compression.c:454:8
>        451
>        452                  folio = filemap_alloc_folio(mapping_gfp_constraint(mapping, constraint_gfp),
>        453                                              0, NULL);
>     -> 454                  if (!folio)
>        455                          break;
> 
> I can reproduce this consistently by running a memory hog concurrently
> with a buffered writer on a machine with a very large amount of swap.
> 
> Commit 7ae37b2c94ed ("btrfs: prevent direct reclaim during compressed
> readahead") clearly intended to suppress these warnings. But because the
> mask set in the address_space with mapping_set_gfp_mask() doesn't include
> __GFP_NOWARN, mapping_gfp_constraint() removes it from constraint_gfp
> before it is passed to filemap_alloc_folio().
> 
> Fix by refactoring the code to add __GFP_NOWARN after the call to
> mapping_gfp_constraint().
> 
> Fixes: 7ae37b2c94ed ("btrfs: prevent direct reclaim during compressed readahead")
> Signed-off-by: Calvin Owens <calvin@wbinvd.org>

This is on me, the original patch https://lore.kernel.org/linux-btrfs/20260323051414.64704-1-jp.kobryn@linux.dev/
had the __GFP_NOWARN at the allocation. As this was not explained as
necessary I moved it to the constant assignments to be grouped. But this
changed the semantics.

Added to for-next, thanks.