| Message ID | 1539592576-24352-1-git-send-email-thunder.leizhen@huawei.com |
|---|---|
| State | New |
| Headers | show |
| Series | [1/1] iommu/arm-smmu-v3: eliminate a potential memory corruption on Hi16xx soc | expand |
On 15/10/2018 09:36, Zhen Lei wrote: > ITS translation register map: > 0x0000-0x003C Reserved > 0x0040 GITS_TRANSLATER > 0x0044-0xFFFC Reserved > Can you add a better opening than the ITS translation register map? > The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon > expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes > data will be written to MSIAddress each time. > > MSIAddr: |----4bytes----|----4bytes----| > | MSIData | IMPDEF | > > There is no problem for ITS, because the next 4 bytes space is reserved in ITS. > But it will overwrite the 4 bytes memory following "sync_count". It's very I think arm_smmu_device.sync_count is better, or "sync_count member in the the smmu driver control struct". > luckly that the previous and the next neighbour of "sync_count" are both aligned /s/luckly/luckily or fortunately/ > by 8 bytes, so no problem is met now. > > It's good to explicitly add a workaround: > 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always > aligned by 8 bytes. > 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. > > There is no functional change. > > Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> > --- > drivers/iommu/arm-smmu-v3.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > index 5059d09..a07bc0d 100644 > --- a/drivers/iommu/arm-smmu-v3.c > +++ b/drivers/iommu/arm-smmu-v3.c > @@ -586,7 +586,10 @@ struct arm_smmu_device { > > struct arm_smmu_strtab_cfg strtab_cfg; > > + union { > + u64 padding; /* workaround for Hisilicon */ I think that a more detailed comment is required. > u32 sync_count; Can you indent these 2 members? However - as discussed internally - this may have endian issue so better to declare full 64b struct. > + } __attribute__((aligned(8))); > > /* IOMMU core code handle */ > struct iommu_device iommu; > Thanks
Hi Zhen, On Mon, Oct 15, 2018 at 04:36:16PM +0800, Zhen Lei wrote: > ITS translation register map: > 0x0000-0x003C Reserved > 0x0040 GITS_TRANSLATER > 0x0044-0xFFFC Reserved > > The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon > expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes > data will be written to MSIAddress each time. > > MSIAddr: |----4bytes----|----4bytes----| > | MSIData | IMPDEF | > > There is no problem for ITS, because the next 4 bytes space is reserved in ITS. > But it will overwrite the 4 bytes memory following "sync_count". It's very > luckly that the previous and the next neighbour of "sync_count" are both aligned > by 8 bytes, so no problem is met now. My understanding is that MSI's are 32bit memory writes and as such the SMMU performs a 32bit write in response to the MSI. If so then what is different with the Hi16xx that causes a problem? Have you been able to able to adjust the layout of the arm_smmu_device struct to demonstrate this? Thanks, Andrew Murray > > It's good to explicitly add a workaround: > 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always > aligned by 8 bytes. > 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. > > There is no functional change. > > Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> > --- > drivers/iommu/arm-smmu-v3.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > index 5059d09..a07bc0d 100644 > --- a/drivers/iommu/arm-smmu-v3.c > +++ b/drivers/iommu/arm-smmu-v3.c > @@ -586,7 +586,10 @@ struct arm_smmu_device { > > struct arm_smmu_strtab_cfg strtab_cfg; > > + union { > + u64 padding; /* workaround for Hisilicon */ > u32 sync_count; > + } __attribute__((aligned(8))); > > /* IOMMU core code handle */ > struct iommu_device iommu; > -- > 1.8.3 > > > _______________________________________________ > iommu mailing list > iommu@lists.linux-foundation.org > https://lists.linuxfoundation.org/mailman/listinfo/iommu
On 15/10/18 09:36, Zhen Lei wrote: > ITS translation register map: > 0x0000-0x003C Reserved > 0x0040 GITS_TRANSLATER > 0x0044-0xFFFC Reserved > > The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon > expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes > data will be written to MSIAddress each time. > > MSIAddr: |----4bytes----|----4bytes----| > | MSIData | IMPDEF | > > There is no problem for ITS, because the next 4 bytes space is reserved in ITS. > But it will overwrite the 4 bytes memory following "sync_count". It's very > luckly that the previous and the next neighbour of "sync_count" are both aligned > by 8 bytes, so no problem is met now. > > It's good to explicitly add a workaround: > 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always > aligned by 8 bytes. > 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. Surely the u64 member inherently makes the union, and thus the u32 member as well, 64-bit-aligned anyway? Robin. > There is no functional change. > > Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> > --- > drivers/iommu/arm-smmu-v3.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > index 5059d09..a07bc0d 100644 > --- a/drivers/iommu/arm-smmu-v3.c > +++ b/drivers/iommu/arm-smmu-v3.c > @@ -586,7 +586,10 @@ struct arm_smmu_device { > > struct arm_smmu_strtab_cfg strtab_cfg; > > + union { > + u64 padding; /* workaround for Hisilicon */ > u32 sync_count; > + } __attribute__((aligned(8))); > > /* IOMMU core code handle */ > struct iommu_device iommu; >
On Mon, Oct 15, 2018 at 04:36:16PM +0800, Zhen Lei wrote: > ITS translation register map: > 0x0000-0x003C Reserved > 0x0040 GITS_TRANSLATER > 0x0044-0xFFFC Reserved > > The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon > expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes > data will be written to MSIAddress each time. > > MSIAddr: |----4bytes----|----4bytes----| > | MSIData | IMPDEF | > > There is no problem for ITS, because the next 4 bytes space is reserved in ITS. > But it will overwrite the 4 bytes memory following "sync_count". It's very > luckly that the previous and the next neighbour of "sync_count" are both aligned > by 8 bytes, so no problem is met now. > > It's good to explicitly add a workaround: > 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always > aligned by 8 bytes. > 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. > > There is no functional change. > > Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> > --- > drivers/iommu/arm-smmu-v3.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > index 5059d09..a07bc0d 100644 > --- a/drivers/iommu/arm-smmu-v3.c > +++ b/drivers/iommu/arm-smmu-v3.c > @@ -586,7 +586,10 @@ struct arm_smmu_device { > > struct arm_smmu_strtab_cfg strtab_cfg; > > + union { > + u64 padding; /* workaround for Hisilicon */ > u32 sync_count; > + } __attribute__((aligned(8))); Won't this already be aligned by the ABI? Anyway, you'll need to swizzle things for big-endian, I suspect. Maybe you can do something clever like making sync_count an array of two elements and determining the offset based on the endianness. Or just keep it simple like we do for things like struct qrwlock and struct qspinlock and use #ifdefs. Also -- you need a comment to explain this insanity :) Will
On 15/10/18 18:21, Will Deacon wrote: > On Mon, Oct 15, 2018 at 04:36:16PM +0800, Zhen Lei wrote: >> ITS translation register map: >> 0x0000-0x003C Reserved >> 0x0040 GITS_TRANSLATER >> 0x0044-0xFFFC Reserved >> >> The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon >> expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes >> data will be written to MSIAddress each time. >> >> MSIAddr: |----4bytes----|----4bytes----| >> | MSIData | IMPDEF | >> >> There is no problem for ITS, because the next 4 bytes space is reserved in ITS. >> But it will overwrite the 4 bytes memory following "sync_count". It's very >> luckly that the previous and the next neighbour of "sync_count" are both aligned >> by 8 bytes, so no problem is met now. >> >> It's good to explicitly add a workaround: >> 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always >> aligned by 8 bytes. >> 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. >> >> There is no functional change. >> >> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> >> --- >> drivers/iommu/arm-smmu-v3.c | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c >> index 5059d09..a07bc0d 100644 >> --- a/drivers/iommu/arm-smmu-v3.c >> +++ b/drivers/iommu/arm-smmu-v3.c >> @@ -586,7 +586,10 @@ struct arm_smmu_device { >> >> struct arm_smmu_strtab_cfg strtab_cfg; >> >> + union { >> + u64 padding; /* workaround for Hisilicon */ >> u32 sync_count; >> + } __attribute__((aligned(8))); > > Won't this already be aligned by the ABI? > > Anyway, you'll need to swizzle things for big-endian, I suspect. Maybe you > can do something clever like making sync_count an array of two elements > and determining the offset based on the endianness. Or just keep it simple > like we do for things like struct qrwlock and struct qspinlock and use > #ifdefs. I don't think so - the CPUs should only ever be making word accesses to the u32 member, while the SMMU expects to be writing little-endian data to an ITS, so AFAICS the data word will always be at the lower address either way. Although now that it's come up, the pre-existing issue of whether the byte order *within* that u32 comes out correct after its round-trip through the SMMU is something I need to run away and hurriedly think about... Robin.
On 2018/10/15 19:17, John Garry wrote: > On 15/10/2018 09:36, Zhen Lei wrote: >> ITS translation register map: >> 0x0000-0x003C Reserved >> 0x0040 GITS_TRANSLATER >> 0x0044-0xFFFC Reserved >> > > Can you add a better opening than the ITS translation register map? OK > >> The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon >> expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes >> data will be written to MSIAddress each time. >> >> MSIAddr: |----4bytes----|----4bytes----| >> | MSIData | IMPDEF | >> >> There is no problem for ITS, because the next 4 bytes space is reserved in ITS. >> But it will overwrite the 4 bytes memory following "sync_count". It's very > > I think arm_smmu_device.sync_count is better, or "sync_count member in the the smmu driver control struct". OK, I will use "struct" in v2. + struct { u32 sync_count; + u32 padding; + } __attribute__((aligned(8))); > >> luckly that the previous and the next neighbour of "sync_count" are both aligned > > /s/luckly/luckily or fortunately/ OK, thanks > >> by 8 bytes, so no problem is met now. >> >> It's good to explicitly add a workaround: >> 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always >> aligned by 8 bytes. >> 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. >> >> There is no functional change. >> >> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> >> --- >> drivers/iommu/arm-smmu-v3.c | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c >> index 5059d09..a07bc0d 100644 >> --- a/drivers/iommu/arm-smmu-v3.c >> +++ b/drivers/iommu/arm-smmu-v3.c >> @@ -586,7 +586,10 @@ struct arm_smmu_device { >> >> struct arm_smmu_strtab_cfg strtab_cfg; >> >> + union { >> + u64 padding; /* workaround for Hisilicon */ > > I think that a more detailed comment is required. OK, I will try to describe it more clearly. > >> u32 sync_count; > > Can you indent these 2 members? However - as discussed internally - this may have endian issue so better to declare full 64b struct. These indent is inherited, to keep aligning with other members. There is no endian issue, I have tested it on both little-endian and big-endian. $gdb vmlinux ...... (gdb) p &((struct arm_smmu_device *)0)->sync_count $1 = (u32 *) 0x4178 (gdb) p &((struct arm_smmu_device *)0)->tst1 $2 = (int *) 0x4170 (gdb) p &((struct arm_smmu_device *)0)->tst2 $3 = (int *) 0x4180 ------------testcase-------- diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c index 5059d09..7c6f7ac 100644 --- a/drivers/iommu/arm-smmu-v3.c +++ b/drivers/iommu/arm-smmu-v3.c @@ -586,7 +586,14 @@ struct arm_smmu_device { struct arm_smmu_strtab_cfg strtab_cfg; + int tst1; + + union { + u64 padding; u32 sync_count; + } __attribute__((aligned(8))); + + int tst2; /* IOMMU core code handle */ struct iommu_device iommu; > >> + } __attribute__((aligned(8))); >> >> /* IOMMU core code handle */ >> struct iommu_device iommu; >> > Thanks > > > > > . > -- Thanks! BestRegards
On 2018/10/15 21:52, Robin Murphy wrote: > On 15/10/18 09:36, Zhen Lei wrote: >> ITS translation register map: >> 0x0000-0x003C Reserved >> 0x0040 GITS_TRANSLATER >> 0x0044-0xFFFC Reserved >> >> The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon >> expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes >> data will be written to MSIAddress each time. >> >> MSIAddr: |----4bytes----|----4bytes----| >> | MSIData | IMPDEF | >> >> There is no problem for ITS, because the next 4 bytes space is reserved in ITS. >> But it will overwrite the 4 bytes memory following "sync_count". It's very >> luckly that the previous and the next neighbour of "sync_count" are both aligned >> by 8 bytes, so no problem is met now. >> >> It's good to explicitly add a workaround: >> 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always >> aligned by 8 bytes. >> 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. > > Surely the u64 member inherently makes the union, and thus the u32 member as well, 64-bit-aligned anyway? Yes, we really only need one step, "1." or "2.". As John Garry suggested, I will change it to "struct" mode. > > Robin. > >> There is no functional change. >> >> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> >> --- >> drivers/iommu/arm-smmu-v3.c | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c >> index 5059d09..a07bc0d 100644 >> --- a/drivers/iommu/arm-smmu-v3.c >> +++ b/drivers/iommu/arm-smmu-v3.c >> @@ -586,7 +586,10 @@ struct arm_smmu_device { >> struct arm_smmu_strtab_cfg strtab_cfg; >> + union { >> + u64 padding; /* workaround for Hisilicon */ >> u32 sync_count; >> + } __attribute__((aligned(8))); >> /* IOMMU core code handle */ >> struct iommu_device iommu; >> > > -- Thanks! BestRegards
On 2018/10/16 1:21, Will Deacon wrote: > On Mon, Oct 15, 2018 at 04:36:16PM +0800, Zhen Lei wrote: >> ITS translation register map: >> 0x0000-0x003C Reserved >> 0x0040 GITS_TRANSLATER >> 0x0044-0xFFFC Reserved >> >> The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon >> expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes >> data will be written to MSIAddress each time. >> >> MSIAddr: |----4bytes----|----4bytes----| >> | MSIData | IMPDEF | >> >> There is no problem for ITS, because the next 4 bytes space is reserved in ITS. >> But it will overwrite the 4 bytes memory following "sync_count". It's very >> luckly that the previous and the next neighbour of "sync_count" are both aligned >> by 8 bytes, so no problem is met now. >> >> It's good to explicitly add a workaround: >> 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always >> aligned by 8 bytes. >> 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. >> >> There is no functional change. >> >> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> >> --- >> drivers/iommu/arm-smmu-v3.c | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c >> index 5059d09..a07bc0d 100644 >> --- a/drivers/iommu/arm-smmu-v3.c >> +++ b/drivers/iommu/arm-smmu-v3.c >> @@ -586,7 +586,10 @@ struct arm_smmu_device { >> >> struct arm_smmu_strtab_cfg strtab_cfg; >> >> + union { >> + u64 padding; /* workaround for Hisilicon */ >> u32 sync_count; >> + } __attribute__((aligned(8))); > > Won't this already be aligned by the ABI? > > Anyway, you'll need to swizzle things for big-endian, I suspect. Maybe you > can do something clever like making sync_count an array of two elements > and determining the offset based on the endianness. Or just keep it simple > like we do for things like struct qrwlock and struct qspinlock and use > #ifdefs. This workaround is a special case, the sync_count is only written by ITS hardware, and is only read by software. Although Hisilicon ITS will write 8 bytes at MSIAddress(required it aligned by 8 bytes), but it can sure that the value of MSIdata will be written at the lower 4 bytes(the start address of sync_count). Because the type of sync_count is u32, so that CPU is also read the 4 bytes at the lower address. > > Also -- you need a comment to explain this insanity :) > > Will > > . > -- Thanks! BestRegards
On 2018/10/15 20:46, Andrew Murray wrote: > Hi Zhen, > > On Mon, Oct 15, 2018 at 04:36:16PM +0800, Zhen Lei wrote: >> ITS translation register map: >> 0x0000-0x003C Reserved >> 0x0040 GITS_TRANSLATER >> 0x0044-0xFFFC Reserved >> >> The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon >> expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes >> data will be written to MSIAddress each time. >> >> MSIAddr: |----4bytes----|----4bytes----| >> | MSIData | IMPDEF | >> >> There is no problem for ITS, because the next 4 bytes space is reserved in ITS. >> But it will overwrite the 4 bytes memory following "sync_count". It's very >> luckly that the previous and the next neighbour of "sync_count" are both aligned >> by 8 bytes, so no problem is met now. > > My understanding is that MSI's are 32bit memory writes and as such the SMMU > performs a 32bit write in response to the MSI. If so then what is different > with the Hi16xx that causes a problem? Have you been able to able to adjust > the layout of the arm_smmu_device struct to demonstrate this? In normal, only 32bits MSIdata will be written into sync_count: |----4bytes----|----4bytes----| | sync_count | xxxx | But for Hi16xx, the ITS hardware will write extra 32bits IMDDEF data into "xxxx". If "xxxx" is the space of the next struct member, its value will be overwritten. > > Thanks, > > Andrew Murray > >> >> It's good to explicitly add a workaround: >> 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always >> aligned by 8 bytes. >> 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. >> >> There is no functional change. >> >> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> >> --- >> drivers/iommu/arm-smmu-v3.c | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c >> index 5059d09..a07bc0d 100644 >> --- a/drivers/iommu/arm-smmu-v3.c >> +++ b/drivers/iommu/arm-smmu-v3.c >> @@ -586,7 +586,10 @@ struct arm_smmu_device { >> >> struct arm_smmu_strtab_cfg strtab_cfg; >> >> + union { >> + u64 padding; /* workaround for Hisilicon */ >> u32 sync_count; >> + } __attribute__((aligned(8))); >> >> /* IOMMU core code handle */ >> struct iommu_device iommu; >> -- >> 1.8.3 >> >> >> _______________________________________________ >> iommu mailing list >> iommu@lists.linux-foundation.org >> https://lists.linuxfoundation.org/mailman/listinfo/iommu > > . > -- Thanks! BestRegards
On Mon, Oct 15, 2018 at 06:36:52PM +0100, Robin Murphy wrote: > On 15/10/18 18:21, Will Deacon wrote: > >On Mon, Oct 15, 2018 at 04:36:16PM +0800, Zhen Lei wrote: > >>ITS translation register map: > >>0x0000-0x003C Reserved > >>0x0040 GITS_TRANSLATER > >>0x0044-0xFFFC Reserved > >> > >>The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon > >>expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes > >>data will be written to MSIAddress each time. > >> > >>MSIAddr: |----4bytes----|----4bytes----| > >> | MSIData | IMPDEF | > >> > >>There is no problem for ITS, because the next 4 bytes space is reserved in ITS. > >>But it will overwrite the 4 bytes memory following "sync_count". It's very > >>luckly that the previous and the next neighbour of "sync_count" are both aligned > >>by 8 bytes, so no problem is met now. > >> > >>It's good to explicitly add a workaround: > >>1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always > >> aligned by 8 bytes. > >>2. Add a "u64" union member to make sure the 4 bytes padding is always exist. > >> > >>There is no functional change. > >> > >>Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> > >>--- > >> drivers/iommu/arm-smmu-v3.c | 3 +++ > >> 1 file changed, 3 insertions(+) > >> > >>diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > >>index 5059d09..a07bc0d 100644 > >>--- a/drivers/iommu/arm-smmu-v3.c > >>+++ b/drivers/iommu/arm-smmu-v3.c > >>@@ -586,7 +586,10 @@ struct arm_smmu_device { > >> struct arm_smmu_strtab_cfg strtab_cfg; > >>+ union { > >>+ u64 padding; /* workaround for Hisilicon */ > >> u32 sync_count; > >>+ } __attribute__((aligned(8))); > > > >Won't this already be aligned by the ABI? > > > >Anyway, you'll need to swizzle things for big-endian, I suspect. Maybe you > >can do something clever like making sync_count an array of two elements > >and determining the offset based on the endianness. Or just keep it simple > >like we do for things like struct qrwlock and struct qspinlock and use > >#ifdefs. > > I don't think so - the CPUs should only ever be making word accesses to the > u32 member, while the SMMU expects to be writing little-endian data to an > ITS, so AFAICS the data word will always be at the lower address either way. Yes, thanks. I'd actually got the union layout wrong in my head and not realised that both member will start at the same address. Ok, so that works. > Although now that it's come up, the pre-existing issue of whether the byte > order *within* that u32 comes out correct after its round-trip through the > SMMU is something I need to run away and hurriedly think about... Ha, oops! Everything going into the command queue is swabbed, so I guess we need to swab the data again in smp_cond_load_acquire(). I'll leave it with you... Will
diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c index 5059d09..a07bc0d 100644 --- a/drivers/iommu/arm-smmu-v3.c +++ b/drivers/iommu/arm-smmu-v3.c @@ -586,7 +586,10 @@ struct arm_smmu_device { struct arm_smmu_strtab_cfg strtab_cfg; + union { + u64 padding; /* workaround for Hisilicon */ u32 sync_count; + } __attribute__((aligned(8))); /* IOMMU core code handle */ struct iommu_device iommu;
ITS translation register map: 0x0000-0x003C Reserved 0x0040 GITS_TRANSLATER 0x0044-0xFFFC Reserved The standard GITS_TRANSLATER register in ITS is only 4 bytes, but Hisilicon expands the next 4 bytes to carry some IMPDEF information. That means, 8 bytes data will be written to MSIAddress each time. MSIAddr: |----4bytes----|----4bytes----| | MSIData | IMPDEF | There is no problem for ITS, because the next 4 bytes space is reserved in ITS. But it will overwrite the 4 bytes memory following "sync_count". It's very luckly that the previous and the next neighbour of "sync_count" are both aligned by 8 bytes, so no problem is met now. It's good to explicitly add a workaround: 1. Add gcc __attribute__((aligned(8))) to make sure that "sync_count" is always aligned by 8 bytes. 2. Add a "u64" union member to make sure the 4 bytes padding is always exist. There is no functional change. Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> --- drivers/iommu/arm-smmu-v3.c | 3 +++ 1 file changed, 3 insertions(+) -- 1.8.3