polkitd segmentation fault on aarch64

[Radha Mohan]

Hi,
In FC21-aarch64 release the polkitd is causing a segmentation fault on
our arm64 platform (Cavium ThunderX). On debugging we found that the
issue is in the mozjs17-17.0.0-14.fc21 package.

The below patch fixed the issue. The max VA bits on ARM64 is 48.

--- a/js/src/jsval.h.old        2015-06-19 14:38:47.000000000 -0400
+++ b/js/src/jsval.h    2015-06-19 17:28:00.653643399 -0400
@@ -188,8 +188,14 @@ typedef uint64_t JSValueShiftedTag;

#elif JS_BITS_PER_WORD == 64

+#if defined(__aarch64__)
+#define JSVAL_PAYLOAD_MASK           0x0000FFFFFFFFFFFFLL
+#define JSVAL_TAG_MASK               0xFFFF000000000000LL
+#else
 #define JSVAL_PAYLOAD_MASK           0x00007FFFFFFFFFFFLL
 #define JSVAL_TAG_MASK               0xFFFF800000000000LL
+#endif
+
 #define JSVAL_TYPE_TO_TAG(type)
((JSValueTag)(JSVAL_TAG_MAX_DOUBLE | (type)))
 #define JSVAL_TYPE_TO_SHIFTED_TAG(type)
(((uint64_t)JSVAL_TYPE_TO_TAG(type)) << JSVAL_TAG_SHIFT)

regards,
Radha Mohan

[Peter Robinson]

Hi Radha,
> In FC21-aarch64 release the polkitd is causing a segmentation fault on
> our arm64 platform (Cavium ThunderX). On debugging we found that the
> issue is in the mozjs17-17.0.0-14.fc21 package.

Can you please report this in bugzilla, and add it a blocker bug of
ARM64 so it could be tracked?

https://bugzilla.redhat.com/enter_bug.cgi?product=Fedora&version=rawhide...

thanks,
Peter

[Radha Mohan]

Please find the filed bug @ https://bugzilla.redhat.com/show_bug.cgi?id=1242326
I mistakenly put the severity as "High", but unable to change it afterwards.

On Sat, Jul 11, 2015 at 4:29 AM, Peter Robinson  wrote:
> Hi Radha,
> > In FC21-aarch64 release the polkitd is causing a segmentation fault on
> > our arm64 platform (Cavium ThunderX). On debugging we found that the
> > issue is in the mozjs17-17.0.0-14.fc21 package.
>
> Can you please report this in bugzilla, and add it a blocker bug of
> ARM64 so it could be tracked?
>
> https://bugzilla.redhat.com/enter_bug.cgi?product=Fedora&version=rawhide...
>
> thanks,
> Peter