percpu: convert spin_lock_irq to spin_lock_irqsave.
authorDennis Zhou <dennis@kernel.org>
Tue, 18 Dec 2018 16:42:27 +0000 (08:42 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 12 Feb 2019 18:47:12 +0000 (19:47 +0100)
[ Upstream commit 6ab7d47bcbf0144a8cb81536c2cead4cde18acfe ]

From Michael Cree:
  "Bisection lead to commit b38d08f3181c ("percpu: restructure
   locking") as being the cause of lockups at initial boot on
   the kernel built for generic Alpha.

   On a suggestion by Tejun Heo that:

   So, the only thing I can think of is that it's calling
   spin_unlock_irq() while irq handling isn't set up yet.
   Can you please try the followings?

   1. Convert all spin_[un]lock_irq() to
      spin_lock_irqsave/unlock_irqrestore()."

Fixes: b38d08f3181c ("percpu: restructure locking")
Reported-and-tested-by: Michael Cree <mcree@orcon.net.nz>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
mm/percpu-km.c

index 38de70a..0f643dc 100644 (file)
@@ -50,6 +50,7 @@ static struct pcpu_chunk *pcpu_create_chunk(gfp_t gfp)
        const int nr_pages = pcpu_group_sizes[0] >> PAGE_SHIFT;
        struct pcpu_chunk *chunk;
        struct page *pages;
+       unsigned long flags;
        int i;
 
        chunk = pcpu_alloc_chunk(gfp);
@@ -68,9 +69,9 @@ static struct pcpu_chunk *pcpu_create_chunk(gfp_t gfp)
        chunk->data = pages;
        chunk->base_addr = page_address(pages) - pcpu_group_offsets[0];
 
-       spin_lock_irq(&pcpu_lock);
+       spin_lock_irqsave(&pcpu_lock, flags);
        pcpu_chunk_populated(chunk, 0, nr_pages, false);
-       spin_unlock_irq(&pcpu_lock);
+       spin_unlock_irqrestore(&pcpu_lock, flags);
 
        pcpu_stats_chunk_alloc();
        trace_percpu_create_chunk(chunk->base_addr);