psr: fix bug which may cause crash
PATCH v2 1/1 psr: fix bug which may cause crash
During test, we found a crash on Xen with below trace. (XEN) Xen call trace:
(XEN) [<ffff82d0802a065a>] R psr.c#l3_cdp_write_msr+0x1e/0x22 (XEN) [<ffff82d0802a0858>] F psr.c#do_write_psr_msrs+0x6d/0x109 (XEN) [<ffff82d08023e000>] F smp_call_function_interrupt+0x5a/0xac (XEN) [<ffff82d0802a2b89>] F call_function_interrupt+0x20/0x34 (XEN) [<ffff82d080282c64>] F do_IRQ+0x175/0x6ae (XEN) [<ffff82d08038b8ba>] F common_interrupt+0x10a/0x120 (XEN) [<ffff82d0802ec616>] F cpu_idle.c#acpi_idle_do_entry+0x9d/0xb1 (XEN) [<ffff82d0802ecc01>] F cpu_idle.c#acpi_processor_idle+0x41d/0x626 (XEN) [<ffff82d08027353b>] F domain.c#idle_loop+0xa5/0xa7
(XEN)
(XEN)
(XEN) ****************************************
(XEN) Panic on CPU 20:
(XEN) GENERAL PROTECTION FAULT
(XEN) [error_code=0000]
(XEN) ****************************************
The bug happens when CDP and MBA co-exist and MBA COS_MAX is bigger than CDP COS_MAX. E.g. MBA has 8 COS registers but CDP only have 6. When setting MBA throttling value for the 7th guest, the value array would be:
+------------------+------------------+--------------+
| Data default val | Code default val | MBA throttle |
+------------------+------------------+--------------+Then, COS id 7 will be selected for writting the values. We should avoid writting CDP data/code valules to COS id 7 MSR because it exceeds the CDP COS_MAX.
--- xen/arch/x86/psr.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
xen/arch/x86/psr.c +2 -1
@@ -1271,7 +1271,8 @@ static void do_write_psr_msrs(void *data)for ( j =0 ; j < cos_num; j++, index++ ) { -if ( feat->cos_reg_val[cos * cos_num + j] != info->val[index] ) +if ( cos <= feat->cos_max && + feat->cos_reg_val[cos * cos_num + j] != info->val[index] ) { feat->cos_reg_val[cos * cos_num + j] = info->val[index]; props->write_msr (cos, info->val[index], props->type[j]);
Signature · 7 lines
1.9.1 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
xen/arch/x86/psr.c static void do_write_psr_msrs(void *data) · line 1275
Quoted 9 lines · Yi Sun
xen/arch/x86/psr.c +2
-1@@ -1271,7 +1271,8 @@ static void do_write_psr_msrs(void *data)for ( j =0 ; j < cos_num; j++, index++ ) { -if ( feat->cos_reg_val[cos * cos_num + j] != info->val[index] ) +if ( cos <= feat->cos_max && + feat->cos_reg_val[cos * cos_num + j] != info->val[index] )
The description is indeed much better now, thanks. However, as indicated in reply to v1, this extra (and at the first glance unmotivated) bounds check wants to be accompanied by a brief but precise comment. Furthermore with the loop bounded by a local variable, why not
cos_num = min(props->cos_num, feat->cos_max + 1);
a few lines up from here (again suitable commented)?
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Quoted 5 lines · Jan Beulich
Quoted 4 lines · Yi Sun
xen/arch/x86/psr.c +0
-0@@ -1271,7 +1271,8 @@ static void do_write_psr_msrs(void *data)
Quoted 9 lines
xen/arch/x86/psr.c static void do_write_psr_msrs(void *data) · line 1275
Quoted 5 lines
xen/arch/x86/psr.c +2
-1for ( j =0 ; j < cos_num; j++, index++ ) { -if ( feat->cos_reg_val[cos * cos_num + j] != info->val[index] ) +if ( cos <= feat->cos_max && + feat->cos_reg_val[cos * cos_num + j] != info->val[index] )The description is indeed much better now, thanks. However, as indicated in reply to v1, this extra (and at the first glance unmotivated) bounds check wants to be accompanied by a brief but
I will add the comment.
Quoted 7 lines
precise comment. Furthermore with the loop bounded by a local variable, why not
cos_num = min(props->cos_num, feat->cos_max + 1);a few lines up from here (again suitable commented)?
cos_num is a different thing with the number of COS registers. The meaning of it is "COS registers number that feature uses for one COS ID". E.g. MBA/CAT cos_num is always 1. But CDP cos_num is 2 because it uses 2 COS registers for one COS ID.
Quoted 1 line
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Quoted 21 lines · Yi Sun
Quoted 5 lines · Jan Beulich
Quoted 4 lines · Yi Sun
xen/arch/x86/psr.c +0
-0@@ -1271,7 +1271,8 @@ static void do_write_psr_msrs(void *data)Quoted 9 lines
xen/arch/x86/psr.c static void do_write_psr_msrs(void *data) · line 1275
Quoted 5 lines
xen/arch/x86/psr.c +2
-1for ( j =0 ; j < cos_num; j++, index++ ) { -if ( feat->cos_reg_val[cos * cos_num + j] != info->val[index] ) +if ( cos <= feat->cos_max && + feat->cos_reg_val[cos * cos_num + j] != info->val[index] )The description is indeed much better now, thanks. However, as indicated in reply to v1, this extra (and at the first glance unmotivated) bounds check wants to be accompanied by a brief but
I will add the comment.
Quoted 3 lines
precise comment. Furthermore with the loop bounded by a local variable, why not
Quoted 8 lines
Quoted 4 lines
cos_num = min(props->cos_num, feat->cos_max + 1);a few lines up from here (again suitable commented)?
cos_numis a different thing with the number of COS registers. The meaning of it is "COS registers number that feature uses for one COS ID". E.g. MBA/CATcos_numis always 1. But CDPcos_numis 2 because it uses 2 COS registers for one COS ID.
Oh, right, I got confused. The cos <= feat->cos_max is entirely loop invariant. Hence the loop shouldn't even be entered, i.e. you want to add a conditional ahead of it.
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel