diff mbox series

target/mips: Fix ll/sc after 7dd547e5ab6b31e7a0cfc182d3ad131dd55a948f

Message ID 20200202153409.28534-1-jrtc27@jrtc27.com (mailing list archive)
State New, archived
Headers show
Series target/mips: Fix ll/sc after 7dd547e5ab6b31e7a0cfc182d3ad131dd55a948f | expand

Commit Message

Jessica Clarke Feb. 2, 2020, 3:34 p.m. UTC
From: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>

After 7dd547e5ab6b31e7a0cfc182d3ad131dd55a948f the env->llval value is
loaded as an unsigned value (instead of sign-extended as before).
Therefore, the CMPXCHG in gen_st_cond() in translate.c fails if the sign
bit is set in the loaded value.
Fix this by sign-extending the llval value for the 32-bit case.

I discovered this issue because FreeBSD MIPS64 was looping forever in an
atomic helper function when trying to start /sbin/init.

Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
Fixes: 7dd547e5ab6b ("target/mips: Use cpu_*_mmuidx_ra instead of MMU_MODE*_SUFFIX")
Buglink: https://bugs.launchpad.net/qemu/+bug/1861605
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: Aleksandar Markovic <amarkovic@wavecomp.com>
Cc: Aleksandar Rikalo <aleksandar.rikalo@rt-rk.com>
Cc: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: James Clarke <jrtc27@jrtc27.com>
---
 target/mips/op_helper.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

--
2.20.1

Comments

Richard Henderson Feb. 3, 2020, 7:52 a.m. UTC | #1
On 2/2/20 3:34 PM, James Clarke wrote:
> From: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
> 
> After 7dd547e5ab6b31e7a0cfc182d3ad131dd55a948f the env->llval value is
> loaded as an unsigned value (instead of sign-extended as before).
> Therefore, the CMPXCHG in gen_st_cond() in translate.c fails if the sign
> bit is set in the loaded value.
> Fix this by sign-extending the llval value for the 32-bit case.
> 
> I discovered this issue because FreeBSD MIPS64 was looping forever in an
> atomic helper function when trying to start /sbin/init.
> 
> Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
> Fixes: 7dd547e5ab6b ("target/mips: Use cpu_*_mmuidx_ra instead of MMU_MODE*_SUFFIX")
> Buglink: https://bugs.launchpad.net/qemu/+bug/1861605
> Cc: Aurelien Jarno <aurelien@aurel32.net>
> Cc: Aleksandar Markovic <amarkovic@wavecomp.com>
> Cc: Aleksandar Rikalo <aleksandar.rikalo@rt-rk.com>
> Cc: Richard Henderson <richard.henderson@linaro.org>
> Signed-off-by: James Clarke <jrtc27@jrtc27.com>
> ---
>  target/mips/op_helper.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)

My fault.  Sorry about that.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~
Philippe Mathieu-Daudé via Feb. 3, 2020, 9:35 a.m. UTC | #2
On Mon, Feb 3, 2020 at 8:52 AM Richard Henderson
<richard.henderson@linaro.org> wrote:
> On 2/2/20 3:34 PM, James Clarke wrote:
> > From: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
> >
> > After 7dd547e5ab6b31e7a0cfc182d3ad131dd55a948f the env->llval value is
> > loaded as an unsigned value (instead of sign-extended as before).
> > Therefore, the CMPXCHG in gen_st_cond() in translate.c fails if the sign
> > bit is set in the loaded value.
> > Fix this by sign-extending the llval value for the 32-bit case.

Thanks!

This fix the problem I spent my Saturday debugging:
https://lists.gnu.org/archive/html/qemu-devel/2020-02/msg00071.html
Too bad we crossed by few hours, you'd have saved my Saturday.

Thomas, I think this also fixes the problem you reported here:
https://www.mail-archive.com/qemu-devel@nongnu.org/msg676025.html

Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>

> >
> > I discovered this issue because FreeBSD MIPS64 was looping forever in an
> > atomic helper function when trying to start /sbin/init.
> >
> > Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
> > Fixes: 7dd547e5ab6b ("target/mips: Use cpu_*_mmuidx_ra instead of MMU_MODE*_SUFFIX")
> > Buglink: https://bugs.launchpad.net/qemu/+bug/1861605
> > Cc: Aurelien Jarno <aurelien@aurel32.net>
> > Cc: Aleksandar Markovic <amarkovic@wavecomp.com>
> > Cc: Aleksandar Rikalo <aleksandar.rikalo@rt-rk.com>
> > Cc: Richard Henderson <richard.henderson@linaro.org>
> > Signed-off-by: James Clarke <jrtc27@jrtc27.com>
> > ---
> >  target/mips/op_helper.c | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
>
> My fault.  Sorry about that.
>
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
>
>
> r~
>
Aleksandar Markovic Feb. 3, 2020, 2:24 p.m. UTC | #3
> From: Richard Henderson <richard.henderson@linaro.org>
> Sent: Monday, February 3, 2020 8:52 AM
> To: James Clarke; qemu-devel@nongnu.org
> Cc: Alex Richardson; Aurelien Jarno; Aleksandar Markovic; Aleksandar Rikalo
> Subject: [EXTERNAL]Re: [PATCH] target/mips: Fix ll/sc after > 7dd547e5ab6b31e7a0cfc182d3ad131dd55a948f
> 
> On 2/2/20 3:34 PM, James Clarke wrote:
> > From: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
> >
> > After 7dd547e5ab6b31e7a0cfc182d3ad131dd55a948f the env->llval value is
> > loaded as an unsigned value (instead of sign-extended as before).
> > Therefore, the CMPXCHG in gen_st_cond() in translate.c fails if the sign
> > bit is set in the loaded value.
> > Fix this by sign-extending the llval value for the 32-bit case.
> >
> > I discovered this issue because FreeBSD MIPS64 was looping forever in an
> > atomic helper function when trying to start /sbin/init.
> >
> > Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
> > Fixes: 7dd547e5ab6b ("target/mips: Use cpu_*_mmuidx_ra instead of MMU_MODE*_SUFFIX")
> > Buglink: https://bugs.launchpad.net/qemu/+bug/1861605
> > Cc: Aurelien Jarno <aurelien@aurel32.net>
> > Cc: Aleksandar Markovic <amarkovic@wavecomp.com>
> > Cc: Aleksandar Rikalo <aleksandar.rikalo@rt-rk.com>
> > Cc: Richard Henderson <richard.henderson@linaro.org>
> > Signed-off-by: James Clarke <jrtc27@jrtc27.com>
> > ---
> >  target/mips/op_helper.c | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> My fault.  Sorry about that.
> 
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> 

Applied to the next mips queue.

Thanks to everyone involved!!

Aleksandar

> 
r~
diff mbox series

Patch

diff --git a/target/mips/op_helper.c b/target/mips/op_helper.c
index 15d05a5fbc..467914d1ca 100644
--- a/target/mips/op_helper.c
+++ b/target/mips/op_helper.c
@@ -305,7 +305,7 @@  static inline hwaddr do_translate_address(CPUMIPSState *env,
     }
 }

-#define HELPER_LD_ATOMIC(name, insn, almask)                                  \
+#define HELPER_LD_ATOMIC(name, insn, almask, do_cast)                         \
 target_ulong helper_##name(CPUMIPSState *env, target_ulong arg, int mem_idx)  \
 {                                                                             \
     if (arg & almask) {                                                       \
@@ -316,12 +316,12 @@  target_ulong helper_##name(CPUMIPSState *env, target_ulong arg, int mem_idx)  \
     }                                                                         \
     env->CP0_LLAddr = do_translate_address(env, arg, 0, GETPC());             \
     env->lladdr = arg;                                                        \
-    env->llval = cpu_##insn##_mmuidx_ra(env, arg, mem_idx, GETPC());          \
+    env->llval = do_cast cpu_##insn##_mmuidx_ra(env, arg, mem_idx, GETPC());  \
     return env->llval;                                                        \
 }
-HELPER_LD_ATOMIC(ll, ldl, 0x3)
+HELPER_LD_ATOMIC(ll, ldl, 0x3, (target_long)(int32_t))
 #ifdef TARGET_MIPS64
-HELPER_LD_ATOMIC(lld, ldq, 0x7)
+HELPER_LD_ATOMIC(lld, ldq, 0x7, (target_ulong))
 #endif
 #undef HELPER_LD_ATOMIC
 #endif