# \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - Subsystems: LINUX FOR POWERPC (32-BIT AND 64-BIT), THE REST - Maintainers: Madhavan Srinivasan, Linus Torvalds - List: linuxppc-dev - 12 messages, 4 authors - Page: https://ratatoskr.run/linuxppc-dev/2016/01/9487572/t ## \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Alexey Kardashevskiy \ - Date: 2016-01-12 04:41:24 - Message: https://ratatoskr.run/linuxppc-dev/2016/01/9487572.md ```text Quite often drivers set only "write" permission assuming that this includes "read" permission as well and this works on plenty platforms. However IODA2 is strict about this and produces an EEH when "read" permission is not and reading happens. This adds a workaround in IODA code to always add the "read" bit when the "write" bit is set. Cc: Benjamin Herrenschmidt Signed-off-by: Alexey Kardashevskiy --- Ben, what was the driver which did not set "read" and caused EEH? --- arch/powerpc/platforms/powernv/pci.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/arch/powerpc/platforms/powernv/pci.c b/arch/powerpc/platforms/powernv/pci.c index f2dd772..c7dcae5 100644 --- a/arch/powerpc/platforms/powernv/pci.c +++ b/arch/powerpc/platforms/powernv/pci.c @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long index, long npages, u64 rpn = __pa(uaddr) >> tbl->it_page_shift; long i; + if (proto_tce & TCE_PCI_WRITE) + proto_tce |= TCE_PCI_READ; + for (i = 0; i < npages; i++) { unsigned long newtce = proto_tce | ((rpn + i) << tbl->it_page_shift); @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long index, BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); + if (newtce & TCE_PCI_WRITE) + newtce |= TCE_PCI_READ; + oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | TCE_PCI_WRITE); *direction = iommu_tce_direction(oldtce); -- 2.5.0.rc3 ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Benjamin Herrenschmidt \ - Date: 2016-01-12 23:07:45 - Message: https://ratatoskr.run/linuxppc-dev/2016/01/9487624.md ```text On Tue, 2016-01-12 at 15:40 +1100, Alexey Kardashevskiy wrote: > Quite often drivers set only "write" permission assuming that this > includes "read" permission as well and this works on plenty > platforms. > However IODA2 is strict about this and produces an EEH when "read" > permission is not and reading happens. > > This adds a workaround in IODA code to always add the "read" bit when > the "write" bit is set. > > Cc: Benjamin Herrenschmidt > Signed-off-by: Alexey Kardashevskiy > --- > > > Ben, what was the driver which did not set "read" and caused EEH? aacraid Cheers, Ben. > > --- >  arch/powerpc/platforms/powernv/pci.c | 6 ++++++ >  1 file changed, 6 insertions(+) > > diff --git a/arch/powerpc/platforms/powernv/pci.c > b/arch/powerpc/platforms/powernv/pci.c > index f2dd772..c7dcae5 100644 > --- a/arch/powerpc/platforms/powernv/pci.c > +++ b/arch/powerpc/platforms/powernv/pci.c > @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long > index, long npages, >   u64 rpn = __pa(uaddr) >> tbl->it_page_shift; >   long i; >   > + if (proto_tce & TCE_PCI_WRITE) > + proto_tce |= TCE_PCI_READ; > + >   for (i = 0; i < npages; i++) { >   unsigned long newtce = proto_tce | >   ((rpn + i) << tbl->it_page_shift); > @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long > index, >   >   BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); >   > + if (newtce & TCE_PCI_WRITE) > + newtce |= TCE_PCI_READ; > + >   oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); >   *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | > TCE_PCI_WRITE); >   *direction = iommu_tce_direction(oldtce); ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Douglas Miller \ - Date: 2016-01-13 02:24:08 - Message: https://ratatoskr.run/linuxppc-dev/2016/01/9487627.md ```text On 01/12/2016 05:07 PM, Benjamin Herrenschmidt wrote: > On Tue, 2016-01-12 at 15:40 +1100, Alexey Kardashevskiy wrote: >> Quite often drivers set only "write" permission assuming that this >> includes "read" permission as well and this works on plenty >> platforms. >> However IODA2 is strict about this and produces an EEH when "read" >> permission is not and reading happens. >> >> This adds a workaround in IODA code to always add the "read" bit when >> the "write" bit is set. >> >> Cc: Benjamin Herrenschmidt >> Signed-off-by: Alexey Kardashevskiy >> --- >> >> >> Ben, what was the driver which did not set "read" and caused EEH? > aacraid > > Cheers, > Ben. Just to be precise, the driver wasn't responsible for setting READ. The driver called scsi_dma_map() and the scsicmd was set (by scsi layer) as DMA_FROM_DEVICE so the current code would set the permissions to WRITE-ONLY. Previously, and in other architectures, this scsicmd would have resulted in READ+WRITE permissions on the DMA map. > >> --- >> arch/powerpc/platforms/powernv/pci.c | 6 ++++++ >> 1 file changed, 6 insertions(+) >> >> diff --git a/arch/powerpc/platforms/powernv/pci.c >> b/arch/powerpc/platforms/powernv/pci.c >> index f2dd772..c7dcae5 100644 >> --- a/arch/powerpc/platforms/powernv/pci.c >> +++ b/arch/powerpc/platforms/powernv/pci.c >> @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long >> index, long npages, >> u64 rpn = __pa(uaddr) >> tbl->it_page_shift; >> long i; >> >> + if (proto_tce & TCE_PCI_WRITE) >> + proto_tce |= TCE_PCI_READ; >> + >> for (i = 0; i < npages; i++) { >> unsigned long newtce = proto_tce | >> ((rpn + i) << tbl->it_page_shift); >> @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long >> index, >> >> BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); >> >> + if (newtce & TCE_PCI_WRITE) >> + newtce |= TCE_PCI_READ; >> + >> oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); >> *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | >> TCE_PCI_WRITE); >> *direction = iommu_tce_direction(oldtce); > _______________________________________________ > Linuxppc-dev mailing list > Linuxppc-dev@lists.ozlabs.org > https://lists.ozlabs.org/listinfo/linuxppc-dev ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Alexey Kardashevskiy \ - Date: 2016-01-19 03:52:25 - Message: https://ratatoskr.run/linuxppc-dev/2016/01/9487695.md ```text On 01/13/2016 01:24 PM, Douglas Miller wrote: > > > On 01/12/2016 05:07 PM, Benjamin Herrenschmidt wrote: >> On Tue, 2016-01-12 at 15:40 +1100, Alexey Kardashevskiy wrote: >>> Quite often drivers set only "write" permission assuming that this >>> includes "read" permission as well and this works on plenty >>> platforms. >>> However IODA2 is strict about this and produces an EEH when "read" >>> permission is not and reading happens. >>> >>> This adds a workaround in IODA code to always add the "read" bit when >>> the "write" bit is set. >>> >>> Cc: Benjamin Herrenschmidt >>> Signed-off-by: Alexey Kardashevskiy >>> --- >>> >>> >>> Ben, what was the driver which did not set "read" and caused EEH? >> aacraid >> >> Cheers, >> Ben. > Just to be precise, the driver wasn't responsible for setting READ. The > driver called scsi_dma_map() and the scsicmd was set (by scsi layer) as > DMA_FROM_DEVICE so the current code would set the permissions to > WRITE-ONLY. Previously, and in other architectures, this scsicmd would have > resulted in READ+WRITE permissions on the DMA map. Does the patch fix the issue? Thanks. >> >>> --- >>> arch/powerpc/platforms/powernv/pci.c | 6 ++++++ >>> 1 file changed, 6 insertions(+) >>> >>> diff --git a/arch/powerpc/platforms/powernv/pci.c >>> b/arch/powerpc/platforms/powernv/pci.c >>> index f2dd772..c7dcae5 100644 >>> --- a/arch/powerpc/platforms/powernv/pci.c >>> +++ b/arch/powerpc/platforms/powernv/pci.c >>> @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long >>> index, long npages, >>> u64 rpn = __pa(uaddr) >> tbl->it_page_shift; >>> long i; >>> + if (proto_tce & TCE_PCI_WRITE) >>> + proto_tce |= TCE_PCI_READ; >>> + >>> for (i = 0; i < npages; i++) { >>> unsigned long newtce = proto_tce | >>> ((rpn + i) << tbl->it_page_shift); >>> @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long >>> index, >>> BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); >>> + if (newtce & TCE_PCI_WRITE) >>> + newtce |= TCE_PCI_READ; >>> + >>> oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); >>> *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | >>> TCE_PCI_WRITE); >>> *direction = iommu_tce_direction(oldtce); >> _______________________________________________ >> Linuxppc-dev mailing list >> Linuxppc-dev@lists.ozlabs.org >> https://lists.ozlabs.org/listinfo/linuxppc-dev > > _______________________________________________ > Linuxppc-dev mailing list > Linuxppc-dev@lists.ozlabs.org > https://lists.ozlabs.org/listinfo/linuxppc-dev -- Alexey ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Douglas Miller \ - Date: 2016-01-19 19:01:29 - Message: https://ratatoskr.run/linuxppc-dev/2016/01/9487697.md ```text On 01/18/2016 09:52 PM, Alexey Kardashevskiy wrote: > On 01/13/2016 01:24 PM, Douglas Miller wrote: >> >> >> On 01/12/2016 05:07 PM, Benjamin Herrenschmidt wrote: >>> On Tue, 2016-01-12 at 15:40 +1100, Alexey Kardashevskiy wrote: >>>> Quite often drivers set only "write" permission assuming that this >>>> includes "read" permission as well and this works on plenty >>>> platforms. >>>> However IODA2 is strict about this and produces an EEH when "read" >>>> permission is not and reading happens. >>>> >>>> This adds a workaround in IODA code to always add the "read" bit when >>>> the "write" bit is set. >>>> >>>> Cc: Benjamin Herrenschmidt >>>> Signed-off-by: Alexey Kardashevskiy >>>> --- >>>> >>>> >>>> Ben, what was the driver which did not set "read" and caused EEH? >>> aacraid >>> >>> Cheers, >>> Ben. >> Just to be precise, the driver wasn't responsible for setting READ. The >> driver called scsi_dma_map() and the scsicmd was set (by scsi layer) as >> DMA_FROM_DEVICE so the current code would set the permissions to >> WRITE-ONLY. Previously, and in other architectures, this scsicmd would have >> resulted in READ+WRITE permissions on the DMA map. > > > Does the patch fix the issue? Thanks. > > > >>> >>>> --- >>>> arch/powerpc/platforms/powernv/pci.c | 6 ++++++ >>>> 1 file changed, 6 insertions(+) >>>> >>>> diff --git a/arch/powerpc/platforms/powernv/pci.c >>>> b/arch/powerpc/platforms/powernv/pci.c >>>> index f2dd772..c7dcae5 100644 >>>> --- a/arch/powerpc/platforms/powernv/pci.c >>>> +++ b/arch/powerpc/platforms/powernv/pci.c >>>> @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long >>>> index, long npages, >>>> u64 rpn = __pa(uaddr) >> tbl->it_page_shift; >>>> long i; >>>> + if (proto_tce & TCE_PCI_WRITE) >>>> + proto_tce |= TCE_PCI_READ; >>>> + >>>> for (i = 0; i < npages; i++) { >>>> unsigned long newtce = proto_tce | >>>> ((rpn + i) << tbl->it_page_shift); >>>> @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long >>>> index, >>>> BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); >>>> + if (newtce & TCE_PCI_WRITE) >>>> + newtce |= TCE_PCI_READ; >>>> + >>>> oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); >>>> *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | >>>> TCE_PCI_WRITE); >>>> *direction = iommu_tce_direction(oldtce); >>> _______________________________________________ >>> Linuxppc-dev mailing list >>> Linuxppc-dev@lists.ozlabs.org >>> https://lists.ozlabs.org/listinfo/linuxppc-dev >> >> _______________________________________________ >> Linuxppc-dev mailing list >> Linuxppc-dev@lists.ozlabs.org >> https://lists.ozlabs.org/listinfo/linuxppc-dev > > I am still working on getting a machine to try this on. From code inspection, it looks like it should work. The problem is shortage of machines and machines tied-up by Test. Thanks, Doug ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Alexey Kardashevskiy \ - Date: 2016-02-09 01:37:26 - Message: https://ratatoskr.run/linuxppc-dev/2016/02/9487990.md ```text On 01/20/2016 06:01 AM, Douglas Miller wrote: > > > On 01/18/2016 09:52 PM, Alexey Kardashevskiy wrote: >> On 01/13/2016 01:24 PM, Douglas Miller wrote: >>> >>> >>> On 01/12/2016 05:07 PM, Benjamin Herrenschmidt wrote: >>>> On Tue, 2016-01-12 at 15:40 +1100, Alexey Kardashevskiy wrote: >>>>> Quite often drivers set only "write" permission assuming that this >>>>> includes "read" permission as well and this works on plenty >>>>> platforms. >>>>> However IODA2 is strict about this and produces an EEH when "read" >>>>> permission is not and reading happens. >>>>> >>>>> This adds a workaround in IODA code to always add the "read" bit when >>>>> the "write" bit is set. >>>>> >>>>> Cc: Benjamin Herrenschmidt >>>>> Signed-off-by: Alexey Kardashevskiy >>>>> --- >>>>> >>>>> >>>>> Ben, what was the driver which did not set "read" and caused EEH? >>>> aacraid >>>> >>>> Cheers, >>>> Ben. >>> Just to be precise, the driver wasn't responsible for setting READ. The >>> driver called scsi_dma_map() and the scsicmd was set (by scsi layer) as >>> DMA_FROM_DEVICE so the current code would set the permissions to >>> WRITE-ONLY. Previously, and in other architectures, this scsicmd would have >>> resulted in READ+WRITE permissions on the DMA map. >> >> >> Does the patch fix the issue? Thanks. >> >> >> >>>> >>>>> --- >>>>> arch/powerpc/platforms/powernv/pci.c | 6 ++++++ >>>>> 1 file changed, 6 insertions(+) >>>>> >>>>> diff --git a/arch/powerpc/platforms/powernv/pci.c >>>>> b/arch/powerpc/platforms/powernv/pci.c >>>>> index f2dd772..c7dcae5 100644 >>>>> --- a/arch/powerpc/platforms/powernv/pci.c >>>>> +++ b/arch/powerpc/platforms/powernv/pci.c >>>>> @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long >>>>> index, long npages, >>>>> u64 rpn = __pa(uaddr) >> tbl->it_page_shift; >>>>> long i; >>>>> + if (proto_tce & TCE_PCI_WRITE) >>>>> + proto_tce |= TCE_PCI_READ; >>>>> + >>>>> for (i = 0; i < npages; i++) { >>>>> unsigned long newtce = proto_tce | >>>>> ((rpn + i) << tbl->it_page_shift); >>>>> @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long >>>>> index, >>>>> BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); >>>>> + if (newtce & TCE_PCI_WRITE) >>>>> + newtce |= TCE_PCI_READ; >>>>> + >>>>> oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); >>>>> *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | >>>>> TCE_PCI_WRITE); >>>>> *direction = iommu_tce_direction(oldtce); >>>> _______________________________________________ >>>> Linuxppc-dev mailing list >>>> Linuxppc-dev@lists.ozlabs.org >>>> https://lists.ozlabs.org/listinfo/linuxppc-dev >>> >>> _______________________________________________ >>> Linuxppc-dev mailing list >>> Linuxppc-dev@lists.ozlabs.org >>> https://lists.ozlabs.org/listinfo/linuxppc-dev >> >> > I am still working on getting a machine to try this on. From code > inspection, it looks like it should work. The problem is shortage of > machines and machines tied-up by Test. Any progress here? Thanks. -- Alexey ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Douglas Miller \ - Date: 2016-02-09 14:28:25 - Message: https://ratatoskr.run/linuxppc-dev/2016/02/9488009.md ```text We finally got the chance to test it end of last week. I forgot to update everyone Monday. B all appearances, the patch fixes the problem. We did not see any new issues with the patch (vs. same test scenarios without). I'll also update the bugzilla. Thanks, Doug On 02/08/2016 07:37 PM, Alexey Kardashevskiy wrote: > On 01/20/2016 06:01 AM, Douglas Miller wrote: >> >> >> On 01/18/2016 09:52 PM, Alexey Kardashevskiy wrote: >>> On 01/13/2016 01:24 PM, Douglas Miller wrote: >>>> >>>> >>>> On 01/12/2016 05:07 PM, Benjamin Herrenschmidt wrote: >>>>> On Tue, 2016-01-12 at 15:40 +1100, Alexey Kardashevskiy wrote: >>>>>> Quite often drivers set only "write" permission assuming that this >>>>>> includes "read" permission as well and this works on plenty >>>>>> platforms. >>>>>> However IODA2 is strict about this and produces an EEH when "read" >>>>>> permission is not and reading happens. >>>>>> >>>>>> This adds a workaround in IODA code to always add the "read" bit when >>>>>> the "write" bit is set. >>>>>> >>>>>> Cc: Benjamin Herrenschmidt >>>>>> Signed-off-by: Alexey Kardashevskiy >>>>>> --- >>>>>> >>>>>> >>>>>> Ben, what was the driver which did not set "read" and caused EEH? >>>>> aacraid >>>>> >>>>> Cheers, >>>>> Ben. >>>> Just to be precise, the driver wasn't responsible for setting READ. The >>>> driver called scsi_dma_map() and the scsicmd was set (by scsi layer) as >>>> DMA_FROM_DEVICE so the current code would set the permissions to >>>> WRITE-ONLY. Previously, and in other architectures, this scsicmd would have >>>> resulted in READ+WRITE permissions on the DMA map. >>> >>> >>> Does the patch fix the issue? Thanks. >>> >>> >>> >>>>> >>>>>> --- >>>>>> arch/powerpc/platforms/powernv/pci.c | 6 ++++++ >>>>>> 1 file changed, 6 insertions(+) >>>>>> >>>>>> diff --git a/arch/powerpc/platforms/powernv/pci.c >>>>>> b/arch/powerpc/platforms/powernv/pci.c >>>>>> index f2dd772..c7dcae5 100644 >>>>>> --- a/arch/powerpc/platforms/powernv/pci.c >>>>>> +++ b/arch/powerpc/platforms/powernv/pci.c >>>>>> @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long >>>>>> index, long npages, >>>>>> u64 rpn = __pa(uaddr) >> tbl->it_page_shift; >>>>>> long i; >>>>>> + if (proto_tce & TCE_PCI_WRITE) >>>>>> + proto_tce |= TCE_PCI_READ; >>>>>> + >>>>>> for (i = 0; i < npages; i++) { >>>>>> unsigned long newtce = proto_tce | >>>>>> ((rpn + i) << tbl->it_page_shift); >>>>>> @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long >>>>>> index, >>>>>> BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); >>>>>> + if (newtce & TCE_PCI_WRITE) >>>>>> + newtce |= TCE_PCI_READ; >>>>>> + >>>>>> oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); >>>>>> *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | >>>>>> TCE_PCI_WRITE); >>>>>> *direction = iommu_tce_direction(oldtce); >>>>> _______________________________________________ >>>>> Linuxppc-dev mailing list >>>>> Linuxppc-dev@lists.ozlabs.org >>>>> https://lists.ozlabs.org/listinfo/linuxppc-dev >>>> >>>> _______________________________________________ >>>> Linuxppc-dev mailing list >>>> Linuxppc-dev@lists.ozlabs.org >>>> https://lists.ozlabs.org/listinfo/linuxppc-dev >>> >>> >> I am still working on getting a machine to try this on. From code >> inspection, it looks like it should work. The problem is shortage of >> machines and machines tied-up by Test. > > Any progress here? Thanks. > > > > ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Alexey Kardashevskiy \ - Date: 2016-02-10 00:32:41 - Message: https://ratatoskr.run/linuxppc-dev/2016/02/9488015.md ```text On 02/10/2016 01:28 AM, Douglas Miller wrote: > We finally got the chance to test it end of last week. I forgot to update > everyone Monday. B all appearances, the patch fixes the problem. We did not > see any new issues with the patch (vs. same test scenarios without). > > I'll also update the bugzilla. Thanks. Care to add "Tested-by"? > > Thanks, > Doug > > On 02/08/2016 07:37 PM, Alexey Kardashevskiy wrote: >> On 01/20/2016 06:01 AM, Douglas Miller wrote: >>> >>> >>> On 01/18/2016 09:52 PM, Alexey Kardashevskiy wrote: >>>> On 01/13/2016 01:24 PM, Douglas Miller wrote: >>>>> >>>>> >>>>> On 01/12/2016 05:07 PM, Benjamin Herrenschmidt wrote: >>>>>> On Tue, 2016-01-12 at 15:40 +1100, Alexey Kardashevskiy wrote: >>>>>>> Quite often drivers set only "write" permission assuming that this >>>>>>> includes "read" permission as well and this works on plenty >>>>>>> platforms. >>>>>>> However IODA2 is strict about this and produces an EEH when "read" >>>>>>> permission is not and reading happens. >>>>>>> >>>>>>> This adds a workaround in IODA code to always add the "read" bit when >>>>>>> the "write" bit is set. >>>>>>> >>>>>>> Cc: Benjamin Herrenschmidt >>>>>>> Signed-off-by: Alexey Kardashevskiy >>>>>>> --- >>>>>>> >>>>>>> >>>>>>> Ben, what was the driver which did not set "read" and caused EEH? >>>>>> aacraid >>>>>> >>>>>> Cheers, >>>>>> Ben. >>>>> Just to be precise, the driver wasn't responsible for setting READ. The >>>>> driver called scsi_dma_map() and the scsicmd was set (by scsi layer) as >>>>> DMA_FROM_DEVICE so the current code would set the permissions to >>>>> WRITE-ONLY. Previously, and in other architectures, this scsicmd would >>>>> have >>>>> resulted in READ+WRITE permissions on the DMA map. >>>> >>>> >>>> Does the patch fix the issue? Thanks. >>>> >>>> >>>> >>>>>> >>>>>>> --- >>>>>>> arch/powerpc/platforms/powernv/pci.c | 6 ++++++ >>>>>>> 1 file changed, 6 insertions(+) >>>>>>> >>>>>>> diff --git a/arch/powerpc/platforms/powernv/pci.c >>>>>>> b/arch/powerpc/platforms/powernv/pci.c >>>>>>> index f2dd772..c7dcae5 100644 >>>>>>> --- a/arch/powerpc/platforms/powernv/pci.c >>>>>>> +++ b/arch/powerpc/platforms/powernv/pci.c >>>>>>> @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long >>>>>>> index, long npages, >>>>>>> u64 rpn = __pa(uaddr) >> tbl->it_page_shift; >>>>>>> long i; >>>>>>> + if (proto_tce & TCE_PCI_WRITE) >>>>>>> + proto_tce |= TCE_PCI_READ; >>>>>>> + >>>>>>> for (i = 0; i < npages; i++) { >>>>>>> unsigned long newtce = proto_tce | >>>>>>> ((rpn + i) << tbl->it_page_shift); >>>>>>> @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long >>>>>>> index, >>>>>>> BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); >>>>>>> + if (newtce & TCE_PCI_WRITE) >>>>>>> + newtce |= TCE_PCI_READ; >>>>>>> + >>>>>>> oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); >>>>>>> *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | >>>>>>> TCE_PCI_WRITE); >>>>>>> *direction = iommu_tce_direction(oldtce); >>>>>> _______________________________________________ >>>>>> Linuxppc-dev mailing list >>>>>> Linuxppc-dev@lists.ozlabs.org >>>>>> https://lists.ozlabs.org/listinfo/linuxppc-dev >>>>> >>>>> _______________________________________________ >>>>> Linuxppc-dev mailing list >>>>> Linuxppc-dev@lists.ozlabs.org >>>>> https://lists.ozlabs.org/listinfo/linuxppc-dev >>>> >>>> >>> I am still working on getting a machine to try this on. From code >>> inspection, it looks like it should work. The problem is shortage of >>> machines and machines tied-up by Test. >> >> Any progress here? Thanks. >> >> >> >> > -- Alexey ``` ## Re: \[RFC PATCH kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Douglas Miller \ - Date: 2016-02-10 12:26:17 - Message: https://ratatoskr.run/linuxppc-dev/2016/02/9488047.md ```text Tested-by: Douglas Miller On 01/11/2016 10:40 PM, Alexey Kardashevskiy wrote: > Quite often drivers set only "write" permission assuming that this > includes "read" permission as well and this works on plenty platforms. > However IODA2 is strict about this and produces an EEH when "read" > permission is not and reading happens. > > This adds a workaround in IODA code to always add the "read" bit when > the "write" bit is set. > > Cc: Benjamin Herrenschmidt > Signed-off-by: Alexey Kardashevskiy > --- > > > Ben, what was the driver which did not set "read" and caused EEH? > > > --- > arch/powerpc/platforms/powernv/pci.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/arch/powerpc/platforms/powernv/pci.c b/arch/powerpc/platforms/powernv/pci.c > index f2dd772..c7dcae5 100644 > --- a/arch/powerpc/platforms/powernv/pci.c > +++ b/arch/powerpc/platforms/powernv/pci.c > @@ -601,6 +601,9 @@ int pnv_tce_build(struct iommu_table *tbl, long index, long npages, > u64 rpn = __pa(uaddr) >> tbl->it_page_shift; > long i; > > + if (proto_tce & TCE_PCI_WRITE) > + proto_tce |= TCE_PCI_READ; > + > for (i = 0; i < npages; i++) { > unsigned long newtce = proto_tce | > ((rpn + i) << tbl->it_page_shift); > @@ -622,6 +625,9 @@ int pnv_tce_xchg(struct iommu_table *tbl, long index, > > BUG_ON(*hpa & ~IOMMU_PAGE_MASK(tbl)); > > + if (newtce & TCE_PCI_WRITE) > + newtce |= TCE_PCI_READ; > + > oldtce = xchg(pnv_tce(tbl, idx), cpu_to_be64(newtce)); > *hpa = be64_to_cpu(oldtce) & ~(TCE_PCI_READ | TCE_PCI_WRITE); > *direction = iommu_tce_direction(oldtce); ``` ## Re: \[RFC, kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Michael Ellerman \ - Date: 2016-02-16 03:18:25 - Message: https://ratatoskr.run/linuxppc-dev/2016/02/9488118.md ```text On Tue, 2016-12-01 at 04:40:20 UTC, Alexey Kardashevskiy wrote: > Quite often drivers set only "write" permission assuming that this > includes "read" permission as well and this works on plenty platforms. > However IODA2 is strict about this and produces an EEH when "read" > permission is not and reading happens. > > This adds a workaround in IODA code to always add the "read" bit when > the "write" bit is set. > > Cc: Benjamin Herrenschmidt > Signed-off-by: Alexey Kardashevskiy > Tested-by: Douglas Miller Are you planning on sending a non-RFC version of this? If so is it an urgent fix I should send upstream now? And if so should it also be CC'ed to stable? cheers ``` ## Re: \[RFC, kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Douglas Miller \ - Date: 2016-02-16 13:20:41 - Message: https://ratatoskr.run/linuxppc-dev/2016/02/9488143.md ```text On 02/15/2016 09:18 PM, Michael Ellerman wrote: > On Tue, 2016-12-01 at 04:40:20 UTC, Alexey Kardashevskiy wrote: >> Quite often drivers set only "write" permission assuming that this >> includes "read" permission as well and this works on plenty platforms. >> However IODA2 is strict about this and produces an EEH when "read" >> permission is not and reading happens. >> >> This adds a workaround in IODA code to always add the "read" bit when >> the "write" bit is set. >> >> Cc: Benjamin Herrenschmidt >> Signed-off-by: Alexey Kardashevskiy >> Tested-by: Douglas Miller > Are you planning on sending a non-RFC version of this? > > If so is it an urgent fix I should send upstream now? And if so should it > also be CC'ed to stable? > > cheers I vote for the more-urgent case. This is the result of a fairly recent change in behavior and we may not have found all the adapters that will break yet. If we get this upstream sooner, it means less exposure. ``` ## Re: \[RFC, kernel\] powerpc/ioda: Set "read" permission when "write" is set - From: Alexey Kardashevskiy \ - Date: 2016-02-17 07:29:53 - Message: https://ratatoskr.run/linuxppc-dev/2016/02/9488182.md ```text On 02/16/2016 02:18 PM, Michael Ellerman wrote: > On Tue, 2016-12-01 at 04:40:20 UTC, Alexey Kardashevskiy wrote: >> Quite often drivers set only "write" permission assuming that this >> includes "read" permission as well and this works on plenty platforms. >> However IODA2 is strict about this and produces an EEH when "read" >> permission is not and reading happens. >> >> This adds a workaround in IODA code to always add the "read" bit when >> the "write" bit is set. >> >> Cc: Benjamin Herrenschmidt >> Signed-off-by: Alexey Kardashevskiy >> Tested-by: Douglas Miller > > Are you planning on sending a non-RFC version of this? Just posted. > If so is it an urgent fix I should send upstream now? Yes. > And if so should it also be CC'ed to stable? Ben suggested that yes, it should. Thanks. Sorry about breaking things. -- Alexey ```