Message ID | 20230922222710.1383808-1-volodymyr_babchuk@epam.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | [v2] ARM: GICv3 ITS: flush caches for newly allocated ITT | expand |
On 9/22/23 18:27, Volodymyr Babchuk wrote: > ITS manages Device Tables and Interrupt Translation Tables on its own, > so generally we are not interested in maintaining any coherence with > CPU's view of those memory regions, except one case: ITS requires that > Interrupt Translation Tables should be initialized with > zeroes. Existing code already does this, but it does not cleans > caches afterwards. This means that ITS may see un-initialized ITT and > CPU can overwrite portions of ITT later, when it finally decides to > flush caches. Visible effect of this issue that there are not > interrupts delivered from a device. > > Fix this by calling clean_and_invalidate_dcache_va_range() for newly > allocated ITT. > > Signed-off-by: Volodymyr Babchuk <volodymyr_babchuk@epam.com> Tested-by: Stewart Hildebrand <stewart.hildebrand@amd.com> > > --- > > Changes since v1: > - Use clean_and_invalidate_dcache_va_range() instead of > clean_dcache_va_range() > - Do this unconditionally > - Do not rename HOST_ITS_FLUSH_CMD_QUEUE into HOST_ITS_FLUSH_BUFFERS > --- > xen/arch/arm/gic-v3-its.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/xen/arch/arm/gic-v3-its.c b/xen/arch/arm/gic-v3-its.c > index 3aa4edda10..8afcd9783b 100644 > --- a/xen/arch/arm/gic-v3-its.c > +++ b/xen/arch/arm/gic-v3-its.c > @@ -685,6 +685,9 @@ int gicv3_its_map_guest_device(struct domain *d, > if ( !itt_addr ) > goto out_unlock; > > + clean_and_invalidate_dcache_va_range(itt_addr, > + nr_events * hw_its->itte_size); > + > dev = xzalloc(struct its_device); > if ( !dev ) > goto out_unlock; > -- > 2.42.0
Hi, (Adding [for-4.18] in the title for Henry to spot the request) On 22/09/2023 23:27, Volodymyr Babchuk wrote: > ITS manages Device Tables and Interrupt Translation Tables on its own, > so generally we are not interested in maintaining any coherence with > CPU's view of those memory regions, except one case: ITS requires that > Interrupt Translation Tables should be initialized with > zeroes. Existing code already does this, but it does not cleans > caches afterwards. This means that ITS may see un-initialized ITT and > CPU can overwrite portions of ITT later, when it finally decides to > flush caches. Visible effect of this issue that there are not > interrupts delivered from a device. > > Fix this by calling clean_and_invalidate_dcache_va_range() for newly > allocated ITT. > I would consider to add: Fixes: 69082e1c210d ("ARM: GICv3 ITS: introduce device mapping") > Signed-off-by: Volodymyr Babchuk <volodymyr_babchuk@epam.com> Reviewed-by: Julien Grall <jgrall@amazon.com> @Henry, this patch should be low-risk. We are cleaning & invalidating the cache, so there should be no change for platform not requiring cache maintenance. This should hopefully had support for more platform. Note that the GICv3 ITS feature is still experimental. Based on what I wrote above, would you be OK to have this patch in 4.18? Cheers,
Hi Julien, Henry, Julien Grall <julien@xen.org> writes: > Hi, > > (Adding [for-4.18] in the title for Henry to spot the request) > > On 22/09/2023 23:27, Volodymyr Babchuk wrote: >> ITS manages Device Tables and Interrupt Translation Tables on its own, >> so generally we are not interested in maintaining any coherence with >> CPU's view of those memory regions, except one case: ITS requires that >> Interrupt Translation Tables should be initialized with >> zeroes. Existing code already does this, but it does not cleans >> caches afterwards. This means that ITS may see un-initialized ITT and >> CPU can overwrite portions of ITT later, when it finally decides to >> flush caches. Visible effect of this issue that there are not >> interrupts delivered from a device. >> Fix this by calling clean_and_invalidate_dcache_va_range() for newly >> allocated ITT. >> > > I would consider to add: > > Fixes: 69082e1c210d ("ARM: GICv3 ITS: introduce device mapping") May I ask you (or Henry?) to add this when you'll commit this change? Or should I publish an updated version? >> Signed-off-by: Volodymyr Babchuk <volodymyr_babchuk@epam.com> > > Reviewed-by: Julien Grall <jgrall@amazon.com> > > @Henry, this patch should be low-risk. We are cleaning & invalidating > the cache, so there should be no change for platform not requiring > cache maintenance. This should hopefully had support for more > platform. Note that the GICv3 ITS feature is still experimental. > > Based on what I wrote above, would you be OK to have this patch in 4.18?
On 25/09/2023 20:00, Volodymyr Babchuk wrote: > > Hi Julien, Henry, > > Julien Grall <julien@xen.org> writes: > >> Hi, >> >> (Adding [for-4.18] in the title for Henry to spot the request) >> >> On 22/09/2023 23:27, Volodymyr Babchuk wrote: >>> ITS manages Device Tables and Interrupt Translation Tables on its own, >>> so generally we are not interested in maintaining any coherence with >>> CPU's view of those memory regions, except one case: ITS requires that >>> Interrupt Translation Tables should be initialized with >>> zeroes. Existing code already does this, but it does not cleans >>> caches afterwards. This means that ITS may see un-initialized ITT and >>> CPU can overwrite portions of ITT later, when it finally decides to >>> flush caches. Visible effect of this issue that there are not >>> interrupts delivered from a device. >>> Fix this by calling clean_and_invalidate_dcache_va_range() for newly >>> allocated ITT. >>> >> >> I would consider to add: >> >> Fixes: 69082e1c210d ("ARM: GICv3 ITS: introduce device mapping") > > May I ask you (or Henry?) to add this when you'll commit this change? Or > should I publish an updated version? I can do it on commit. Cheers,
Hi Julien, > On Sep 26, 2023, at 02:33, Julien Grall <julien@xen.org> wrote: > > Hi, > > (Adding [for-4.18] in the title for Henry to spot the request) Thanks! > > On 22/09/2023 23:27, Volodymyr Babchuk wrote: >> ITS manages Device Tables and Interrupt Translation Tables on its own, >> so generally we are not interested in maintaining any coherence with >> CPU's view of those memory regions, except one case: ITS requires that >> Interrupt Translation Tables should be initialized with >> zeroes. Existing code already does this, but it does not cleans >> caches afterwards. This means that ITS may see un-initialized ITT and >> CPU can overwrite portions of ITT later, when it finally decides to >> flush caches. Visible effect of this issue that there are not >> interrupts delivered from a device. >> Fix this by calling clean_and_invalidate_dcache_va_range() for newly >> allocated ITT. > > I would consider to add: > > Fixes: 69082e1c210d ("ARM: GICv3 ITS: introduce device mapping") > >> Signed-off-by: Volodymyr Babchuk <volodymyr_babchuk@epam.com> > > Reviewed-by: Julien Grall <jgrall@amazon.com> > > @Henry, this patch should be low-risk. We are cleaning & invalidating the cache, so there should be no change for platform not requiring cache maintenance. This should hopefully had support for more platform. Note that the GICv3 ITS feature is still experimental. > > Based on what I wrote above, would you be OK to have this patch in 4.18? Yes, I was about to ask the same question but somehow forgot it. This is a quite low risk patch and I think it is fine to have this in 4.18, so if the "Fixes” tag can be added on commit, please also add: Release-acked-by: Henry Wang <Henry.Wang@arm.com> Kind regards, Henry > > Cheers, > > -- > Julien Grall
On 25/09/2023 23:19, Henry Wang wrote: > Hi Julien, Hi Henry, > Yes, I was about to ask the same question but somehow forgot it. This is a quite > low risk patch and I think it is fine to have this in 4.18, so if the "Fixes” tag > can be added on commit, please also add: > > Release-acked-by: Henry Wang <Henry.Wang@arm.com> Thanks! It is now committed. Cheers,
diff --git a/xen/arch/arm/gic-v3-its.c b/xen/arch/arm/gic-v3-its.c index 3aa4edda10..8afcd9783b 100644 --- a/xen/arch/arm/gic-v3-its.c +++ b/xen/arch/arm/gic-v3-its.c @@ -685,6 +685,9 @@ int gicv3_its_map_guest_device(struct domain *d, if ( !itt_addr ) goto out_unlock; + clean_and_invalidate_dcache_va_range(itt_addr, + nr_events * hw_its->itte_size); + dev = xzalloc(struct its_device); if ( !dev ) goto out_unlock;
ITS manages Device Tables and Interrupt Translation Tables on its own, so generally we are not interested in maintaining any coherence with CPU's view of those memory regions, except one case: ITS requires that Interrupt Translation Tables should be initialized with zeroes. Existing code already does this, but it does not cleans caches afterwards. This means that ITS may see un-initialized ITT and CPU can overwrite portions of ITT later, when it finally decides to flush caches. Visible effect of this issue that there are not interrupts delivered from a device. Fix this by calling clean_and_invalidate_dcache_va_range() for newly allocated ITT. Signed-off-by: Volodymyr Babchuk <volodymyr_babchuk@epam.com> --- Changes since v1: - Use clean_and_invalidate_dcache_va_range() instead of clean_dcache_va_range() - Do this unconditionally - Do not rename HOST_ITS_FLUSH_CMD_QUEUE into HOST_ITS_FLUSH_BUFFERS --- xen/arch/arm/gic-v3-its.c | 3 +++ 1 file changed, 3 insertions(+)