MIME-Version: 1.0
References: <cover.1675111415.git.andreyknvl@google.com> <19512bb03eed27ced5abeb5bd03f9a8381742cb1.1675111415.git.andreyknvl@google.com>
In-Reply-To: <19512bb03eed27ced5abeb5bd03f9a8381742cb1.1675111415.git.andreyknvl@google.com>
From:   Marco Elver <elver@google.com>
Date:   Tue, 31 Jan 2023 09:40:36 +0100
Message-ID: <CANpmjNNzNSDrxfrZUcRtt7=hV=Mz8_kyCpqVnyAqzhaiyipXCg@mail.gmail.com>
Subject: Re: [PATCH 16/18] lib/stackdepot: annotate racy slab_index accesses
To:     andrey.konovalov@linux.dev
Cc:     Alexander Potapenko <glider@google.com>,
        Andrey Konovalov <andreyknvl@gmail.com>,
        Vlastimil Babka <vbabka@suse.cz>, kasan-dev@googlegroups.com,
        Evgenii Stepanov <eugenis@google.com>,
        Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org,
        Andrey Konovalov <andreyknvl@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, 30 Jan 2023 at 21:51, <andrey.konovalov@linux.dev> wrote:
>
> From: Andrey Konovalov <andreyknvl@google.com>
>
> Accesses to slab_index are protected by slab_lock everywhere except
> in a sanity check in stack_depot_fetch. The read access there can race
> with the write access in depot_alloc_stack.
>
> Use WRITE/READ_ONCE() to annotate the racy accesses.
>
> As the sanity check is only used to print a warning in case of a
> violation of the stack depot interface usage, it does not make a lot
> of sense to use proper synchronization.
>
> Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
> ---
>  lib/stackdepot.c | 13 +++++++++----
>  1 file changed, 9 insertions(+), 4 deletions(-)
>
> diff --git a/lib/stackdepot.c b/lib/stackdepot.c
> index f291ad6a4e72..cc2fe8563af4 100644
> --- a/lib/stackdepot.c
> +++ b/lib/stackdepot.c
> @@ -269,8 +269,11 @@ depot_alloc_stack(unsigned long *entries, int size, u32 hash, void **prealloc)
>                         return NULL;
>                 }
>
> -               /* Move on to the next slab. */
> -               slab_index++;
> +               /*
> +                * Move on to the next slab.
> +                * WRITE_ONCE annotates a race with stack_depot_fetch.

"Pairs with potential concurrent read in stack_depot_fetch()." would be clearer.

I wouldn't say WRITE_ONCE annotates a race (race = involves 2+
accesses, but here's just 1), it just marks this access here which
itself is paired with the potential racing read in the other function.

> +                */
> +               WRITE_ONCE(slab_index, slab_index + 1);
>                 slab_offset = 0;
>                 /*
>                  * smp_store_release() here pairs with smp_load_acquire() in
> @@ -492,6 +495,8 @@ unsigned int stack_depot_fetch(depot_stack_handle_t handle,
>                                unsigned long **entries)
>  {
>         union handle_parts parts = { .handle = handle };
> +       /* READ_ONCE annotates a race with depot_alloc_stack. */
> +       int slab_index_cached = READ_ONCE(slab_index);
>         void *slab;
>         size_t offset = parts.offset << DEPOT_STACK_ALIGN;
>         struct stack_record *stack;
> @@ -500,9 +505,9 @@ unsigned int stack_depot_fetch(depot_stack_handle_t handle,
>         if (!handle)
>                 return 0;
>
> -       if (parts.slab_index > slab_index) {
> +       if (parts.slab_index > slab_index_cached) {
>                 WARN(1, "slab index %d out of bounds (%d) for stack id %08x\n",
> -                       parts.slab_index, slab_index, handle);
> +                       parts.slab_index, slab_index_cached, handle);
>                 return 0;
>         }
>         slab = stack_slabs[parts.slab_index];
> --
> 2.25.1
>