Re: [linux-next:master] [memcg] 01d37228d3: netperf.Throughput_Mbps 37.9% regression

[Vlastimil Babka <vbabka@suse.cz>]

On 3/10/25 11:56, Alexei Starovoitov wrote:
> On Mon, Mar 10, 2025 at 11:34 AM Vlastimil Babka <vbabka@suse.cz> wrote:
>>
>> On 3/10/25 11:18, Alexei Starovoitov wrote:
>> >> because this will affect the refill even if consume_stock() fails not due to
>> >> a trylock failure (which should not be happening), but also just because the
>> >> stock was of a wrong memcg or depleted. So in the nowait context we deny the
>> >> refill even if we have the memory. Attached patch could be used to see if it
>> >> if fixes things. I'm not sure about the testcases where it doesn't look like
>> >> nowait context would be used though, let's see.
>> >
>> > Not quite.
>> > GFP_NOWAIT includes __GFP_KSWAPD_RECLAIM,
>> > so gfpflags_allow_spinning() will return true.
>>
>> Uh right, it's the new gfpflags_allow_spinning(), not the
>> gfpflags_allow_blocking() I'm used to and implicitly assumed, sorry.
>>
>> But then it's very simple because it has a bug:
>> gfpflags_allow_spinning() does
>>
>> return !(gfp_flags & __GFP_RECLAIM);
>>
>> should be !!
> 
> Ouch.
> So I accidentally exposed the whole linux-next to this stress testing
> of new trylock facilities :(
> But the silver lining is that this is the only thing that blew up :)
> Could you send a patch or I will do it later today.

OK
----8<----
From 69b3d1631645c82d9d88f17fb01184d24034df2b Mon Sep 17 00:00:00 2001
From: Vlastimil Babka <vbabka@suse.cz>
Date: Mon, 10 Mar 2025 11:57:52 +0100
Subject: [PATCH] mm: Fix the flipped condition in gfpflags_allow_spinning()

The function gfpflags_allow_spinning() has a bug that makes it return
the opposite result than intended. This could contribute to deadlocks as
usage profilerates, for now it was noticed as a performance regression
due to try_charge_memcg() not refilling memcg stock when it could. Fix
the flipped condition.

Fixes: 97769a53f117 ("mm, bpf: Introduce try_alloc_pages() for opportunistic page allocation")
Reported-by: kernel test robot <oliver.sang@intel.com>
Closes: https://lore.kernel.org/oe-lkp/202503101254.cfd454df-lkp@intel.com

Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
---
 include/linux/gfp.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index ceb226c2e25c..c9fa6309c903 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -55,7 +55,7 @@ static inline bool gfpflags_allow_spinning(const gfp_t gfp_flags)
 	 * regular page allocator doesn't fully support this
 	 * allocation mode.
 	 */
-	return !(gfp_flags & __GFP_RECLAIM);
+	return !!(gfp_flags & __GFP_RECLAIM);
 }
 
 #ifdef CONFIG_HIGHMEM
-- 
2.48.1