diff mbox series

[4/4] random: vDSO: don't use 64 bits atomics on 32 bits architectures

Message ID 30806cb8d7e0b95dcfb9f81a4583759faa1d8f31.1724743492.git.christophe.leroy@csgroup.eu (mailing list archive)
State Handled Elsewhere
Headers show
Series Fixups for random vDSO | expand

Commit Message

Christophe Leroy Aug. 27, 2024, 7:31 a.m. UTC
Performing SMP atomic operations on u64 fails on powerpc32:

    CC      drivers/char/random.o
  In file included from <command-line>:
  drivers/char/random.c: In function 'crng_reseed':
  ././include/linux/compiler_types.h:510:45: error: call to '__compiletime_assert_391' declared with attribute error: Need native word sized stores/loads for atomicity.
    510 |         _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
        |                                             ^
  ././include/linux/compiler_types.h:491:25: note: in definition of macro '__compiletime_assert'
    491 |                         prefix ## suffix();                             \
        |                         ^~~~~~
  ././include/linux/compiler_types.h:510:9: note: in expansion of macro '_compiletime_assert'
    510 |         _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
        |         ^~~~~~~~~~~~~~~~~~~
  ././include/linux/compiler_types.h:513:9: note: in expansion of macro 'compiletime_assert'
    513 |         compiletime_assert(__native_word(t),                            \
        |         ^~~~~~~~~~~~~~~~~~
  ./arch/powerpc/include/asm/barrier.h:74:9: note: in expansion of macro 'compiletime_assert_atomic_type'
     74 |         compiletime_assert_atomic_type(*p);                             \
        |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  ./include/asm-generic/barrier.h:172:55: note: in expansion of macro '__smp_store_release'
    172 | #define smp_store_release(p, v) do { kcsan_release(); __smp_store_release(p, v); } while (0)
        |                                                       ^~~~~~~~~~~~~~~~~~~
  drivers/char/random.c:286:9: note: in expansion of macro 'smp_store_release'
    286 |         smp_store_release(&__arch_get_k_vdso_rng_data()->generation, next_gen + 1);
        |         ^~~~~~~~~~~~~~~~~

Random driver generation is handled as unsigned long not u64, see for
instance base_cnrg or struct crng.

But on vDSO it needs to be an u64 not just an unsigned long because of
32 bits VDSO being used with 64 bits kernels.

On random side however it is an unsigned long hence a 32 bits value on
32 bits architectures, so just cast it to unsigned long for the
smp_store_release(). A side effect is that on big endian architectures
the store will be performed in the upper 32 bits. It is not an issue
on its own because the vDSO site doesn't mind the value, it only
checks differences. Just make sure that the vDSO side checks the full
64 bits, for that the local current_generation has to be u64 as well.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
---
v3: Cast to unsigned long in random and use u64 in vDSO instead of changing generation field to unsigned long
---
 drivers/char/random.c | 9 ++++++++-
 lib/vdso/getrandom.c  | 2 +-
 2 files changed, 9 insertions(+), 2 deletions(-)

Comments

Jason A. Donenfeld Aug. 27, 2024, 8:03 a.m. UTC | #1
On Tue, Aug 27, 2024 at 09:31:50AM +0200, Christophe Leroy wrote:
> Performing SMP atomic operations on u64 fails on powerpc32:

Thanks for this, and nice catch on the vDSO side checking on big endian.
I've applied this, fixing up the commit message and the comment,
maintaining the reverse christmas tree in getrandom.c, and adding tglx's
suggested-by tag.

Jason
diff mbox series

Patch

diff --git a/drivers/char/random.c b/drivers/char/random.c
index 77968309e2c2..dc9bab51e74d 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -282,8 +282,15 @@  static void crng_reseed(struct work_struct *work)
 	 * former to arrive at the latter. Use smp_store_release so that this
 	 * is ordered with the write above to base_crng.generation. Pairs with
 	 * the smp_rmb() before the syscall in the vDSO code.
+	 *
+	 * Cast to unsigned long for 32 bits architectures as atomic 64 bits
+	 * operations are not supported on those architectures. Anyway
+	 * base_crng.generation is a 32 bits value so it is ok. On big endian
+	 * architectures it will be stored in the upper 32 bits but that's ok
+	 * because the vDSO side only checks whether the value changed, it
+	 * doesn't use or interpret the value.
 	 */
-	smp_store_release(&__arch_get_k_vdso_rng_data()->generation, next_gen + 1);
+	smp_store_release((unsigned long *)&__arch_get_k_vdso_rng_data()->generation, next_gen + 1);
 #endif
 	if (!static_branch_likely(&crng_is_ready))
 		crng_init = CRNG_READY;
diff --git a/lib/vdso/getrandom.c b/lib/vdso/getrandom.c
index 5d79663b026b..8027b2711b69 100644
--- a/lib/vdso/getrandom.c
+++ b/lib/vdso/getrandom.c
@@ -69,7 +69,7 @@  __cvdso_getrandom_data(const struct vdso_rng_data *rng_info, void *buffer, size_
 	struct vgetrandom_state *state = opaque_state;
 	size_t batch_len, nblocks, orig_len = len;
 	bool in_use, have_retried = false;
-	unsigned long current_generation;
+	u64 current_generation;
 	void *orig_buffer = buffer;
 	u32 counter[2] = { 0 };