Re: [PATCH] riscv: compat: Using seperated vdso_maps for compat_vdso_info

From: Palmer Dabbelt
Date: Wed May 25 2022 - 20:39:49 EST


On Wed, 25 May 2022 15:56:07 PDT (-0700), linux@xxxxxxxxxxxx wrote:
On 5/25/22 14:34, Palmer Dabbelt wrote:
On Wed, 25 May 2022 14:15:03 PDT (-0700), linux@xxxxxxxxxxxx wrote:
On 5/25/22 09:04, guoren@xxxxxxxxxx wrote:
From: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>

This is a fixup for vdso implementation which caused musl to
fail.

[   11.600082] Run /sbin/init as init process
[   11.628561] init[1]: unhandled signal 11 code 0x1 at
0x0000000000000000 in libc.so[ffffff8ad39000+a4000]
[   11.629398] CPU: 0 PID: 1 Comm: init Not tainted
5.18.0-rc7-next-20220520 #1
[   11.629462] Hardware name: riscv-virtio,qemu (DT)
[   11.629546] epc : 00ffffff8ada1100 ra : 00ffffff8ada13c8 sp :
00ffffffc58199f0
[   11.629586]  gp : 00ffffff8ad39000 tp : 00ffffff8ade0998 t0 :
ffffffffffffffff
[   11.629598]  t1 : 00ffffffc5819fd0 t2 : 0000000000000000 s0 :
00ffffff8ade0cc0
[   11.629610]  s1 : 00ffffff8ade0cc0 a0 : 0000000000000000 a1 :
00ffffffc5819a00
[   11.629622]  a2 : 0000000000000001 a3 : 000000000000001e a4 :
00ffffffc5819b00
[   11.629634]  a5 : 00ffffffc5819b00 a6 : 0000000000000000 a7 :
0000000000000000
[   11.629645]  s2 : 00ffffff8ade0ac8 s3 : 00ffffff8ade0ec8 s4 :
00ffffff8ade0728
[   11.629656]  s5 : 00ffffff8ade0a90 s6 : 0000000000000000 s7 :
00ffffffc5819e40
[   11.629667]  s8 : 00ffffff8ade0ca0 s9 : 00ffffff8addba50 s10:
0000000000000000
[   11.629678]  s11: 0000000000000000 t3 : 0000000000000002 t4 :
0000000000000001
[   11.629688]  t5 : 0000000000020000 t6 : ffffffffffffffff
[   11.629699] status: 0000000000004020 badaddr: 0000000000000000
cause: 000000000000000d

The last __vdso_init(&compat_vdso_info) replaces the data in normal
vdso_info. This is an obvious bug.

Reported-by: Guenter Roeck <linux@xxxxxxxxxxxx>
Signed-off-by: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>
Signed-off-by: Guo Ren <guoren@xxxxxxxxxx>
Cc: Palmer Dabbelt <palmer@xxxxxxxxxxx>
Cc: Heiko Stübner <heiko@xxxxxxxxx>

Tested-by: Guenter Roeck <linux@xxxxxxxxxxxx>

Sorry I'm a bit buried right now, this is fixing the issue you pointed out earlier?

Yes.

Awosome, I think that was the only big blocker so far.

I added a musl-based userspace to my test setup, which is rv64-only (buildroot doesn't have rv32 musl, I thought upstream had it but maybe i'm misremembering). This patch fixes the bug, so I've added it to for-next with

Fixes: 3092eb456375 ("riscv: compat: vdso: Add setup additional pages implementation")

which I think is correct, but LMK if there's an issue.
Thanks!

(and also sorry I poked Geert instead of you about this one, oops ;))


Thanks,
Guenter


---
  arch/riscv/kernel/vdso.c | 15 +++++++++++++--
  1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/arch/riscv/kernel/vdso.c b/arch/riscv/kernel/vdso.c
index 50fe4c877603..69b05b6c181b 100644
--- a/arch/riscv/kernel/vdso.c
+++ b/arch/riscv/kernel/vdso.c
@@ -206,12 +206,23 @@ static struct __vdso_info vdso_info __ro_after_init = {
  };

  #ifdef CONFIG_COMPAT
+static struct vm_special_mapping rv_compat_vdso_maps[] __ro_after_init = {
+    [RV_VDSO_MAP_VVAR] = {
+        .name   = "[vvar]",
+        .fault = vvar_fault,
+    },
+    [RV_VDSO_MAP_VDSO] = {
+        .name   = "[vdso]",
+        .mremap = vdso_mremap,
+    },
+};
+
  static struct __vdso_info compat_vdso_info __ro_after_init = {
      .name = "compat_vdso",
      .vdso_code_start = compat_vdso_start,
      .vdso_code_end = compat_vdso_end,
-    .dm = &rv_vdso_maps[RV_VDSO_MAP_VVAR],
-    .cm = &rv_vdso_maps[RV_VDSO_MAP_VDSO],
+    .dm = &rv_compat_vdso_maps[RV_VDSO_MAP_VVAR],
+    .cm = &rv_compat_vdso_maps[RV_VDSO_MAP_VDSO],
  };
  #endif