Zhang, Xiantao
2010-Jan-26 14:10 UTC
[Xen-devel] [PATCH] enhancement for unmaskable MSI and two cosmetic fixes.
0001: Revert 20334, since it doesn''t solve the race for all corner cases. 0002: Fixed unmaskable MSI''s IRQ migration. 0003: Reduce eoi stack''s size to save per-cpu area. 0004: polarity-switch method for EOI message is only used for non-directed-eoi case. Xiantao _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser
2010-Jan-26 14:26 UTC
[Xen-devel] Re: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes.
Are these essential for 4.0.0? I''ve just tagged -rc2, so they''ve missed that. -- Keir On 26/01/2010 14:10, "Zhang, Xiantao" <xiantao.zhang@intel.com> wrote:> 0001: Revert 20334, since it doesn''t solve the race for all corner cases. > 0002: Fixed unmaskable MSI''s IRQ migration. > 0003: Reduce eoi stack''s size to save per-cpu area. > 0004: polarity-switch method for EOI message is only used for non-directed-eoi > case. > Xiantao_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Zhang, Xiantao
2010-Jan-26 14:35 UTC
[Xen-devel] RE: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes.
Yeah, at least we have to apply the first two to fix a bug related to unmaskable MSI. Xiantao -----Original Message----- From: Keir Fraser [mailto:keir.fraser@eu.citrix.com] Sent: Tuesday, January 26, 2010 10:26 PM To: Zhang, Xiantao; xen-devel@lists.xensource.com Subject: Re: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes. Are these essential for 4.0.0? I''ve just tagged -rc2, so they''ve missed that. -- Keir On 26/01/2010 14:10, "Zhang, Xiantao" <xiantao.zhang@intel.com> wrote:> 0001: Revert 20334, since it doesn''t solve the race for all corner cases. > 0002: Fixed unmaskable MSI''s IRQ migration. > 0003: Reduce eoi stack''s size to save per-cpu area. > 0004: polarity-switch method for EOI message is only used for non-directed-eoi > case. > Xiantao_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser
2010-Jan-26 15:01 UTC
[Xen-devel] Re: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes.
Is patch #3 even safe? Looks like if a CPU somehow got all dynamic vectors stacked up, it would overflow into the stackpointer at the end of the array. -- Keir On 26/01/2010 14:35, "Zhang, Xiantao" <xiantao.zhang@intel.com> wrote:> Yeah, at least we have to apply the first two to fix a bug related to > unmaskable MSI. > Xiantao > > -----Original Message----- > From: Keir Fraser [mailto:keir.fraser@eu.citrix.com] > Sent: Tuesday, January 26, 2010 10:26 PM > To: Zhang, Xiantao; xen-devel@lists.xensource.com > Subject: Re: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes. > > Are these essential for 4.0.0? I''ve just tagged -rc2, so they''ve missed > that. > > -- Keir > > On 26/01/2010 14:10, "Zhang, Xiantao" <xiantao.zhang@intel.com> wrote: > >> 0001: Revert 20334, since it doesn''t solve the race for all corner cases. >> 0002: Fixed unmaskable MSI''s IRQ migration. >> 0003: Reduce eoi stack''s size to save per-cpu area. >> 0004: polarity-switch method for EOI message is only used for >> non-directed-eoi >> case. >> Xiantao > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Zhang, Xiantao
2010-Jan-26 15:23 UTC
[Xen-devel] RE: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes.
No, it is also safe even so .The dynamic vector IRQ_MOVE_CLEANUP_VECTOR is used for cleanup vector when IRQ migration and its EOI is never stacked, so only NR_DYNAMIC_VECTORS -1 is really used, and the last array cell is always safe to store stack pointer. Xiantao -----Original Message----- From: Keir Fraser [mailto:keir.fraser@eu.citrix.com] Sent: Tuesday, January 26, 2010 11:02 PM To: Zhang, Xiantao; xen-devel@lists.xensource.com Subject: Re: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes. Is patch #3 even safe? Looks like if a CPU somehow got all dynamic vectors stacked up, it would overflow into the stackpointer at the end of the array. -- Keir On 26/01/2010 14:35, "Zhang, Xiantao" <xiantao.zhang@intel.com> wrote:> Yeah, at least we have to apply the first two to fix a bug related to > unmaskable MSI. > Xiantao > > -----Original Message----- > From: Keir Fraser [mailto:keir.fraser@eu.citrix.com] > Sent: Tuesday, January 26, 2010 10:26 PM > To: Zhang, Xiantao; xen-devel@lists.xensource.com > Subject: Re: [PATCH] enhancement for unmaskable MSI and two cosmetic fixes. > > Are these essential for 4.0.0? I''ve just tagged -rc2, so they''ve missed > that. > > -- Keir > > On 26/01/2010 14:10, "Zhang, Xiantao" <xiantao.zhang@intel.com> wrote: > >> 0001: Revert 20334, since it doesn''t solve the race for all corner cases. >> 0002: Fixed unmaskable MSI''s IRQ migration. >> 0003: Reduce eoi stack''s size to save per-cpu area. >> 0004: polarity-switch method for EOI message is only used for >> non-directed-eoi >> case. >> Xiantao > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel