Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for Substream IDs
On 2017/11/3 17:37, Jean-Philippe Brucker wrote: > On 03/11/17 05:45, Yisheng Xie wrote: >> Hi Jean, >> >> On 2017/11/3 1:02, Shameerali Kolothum Thodi wrote: >>> >>> >>>> -Original Message- >>>> From: Jean-Philippe Brucker [mailto:jean-philippe.bruc...@arm.com] >>>> Sent: Thursday, November 02, 2017 3:52 PM >>>> To: Shameerali Kolothum Thodi <shameerali.kolothum.th...@huawei.com> >>>> Cc: linux-arm-ker...@lists.infradead.org; linux-...@vger.kernel.org; linux- >>>> a...@vger.kernel.org; devicet...@vger.kernel.org; iommu@lists.linux- >>>> foundation.org; Mark Rutland <mark.rutl...@arm.com>; xieyisheng (A) >>>> <xieyishe...@huawei.com>; Gabriele Paoloni >>>> <gabriele.paol...@huawei.com>; Catalin Marinas >>>> <catalin.mari...@arm.com>; Will Deacon <will.dea...@arm.com>; >>>> ok...@codeaurora.org; yi.l@intel.com; Lorenzo Pieralisi >>>> <lorenzo.pieral...@arm.com>; ashok@intel.com; t...@semihalf.com; >>>> j...@8bytes.org; rfr...@cavium.com; l...@kernel.org; >>>> jacob.jun@linux.intel.com; alex.william...@redhat.com; >>>> robh...@kernel.org; Leizhen (ThunderTown) <thunder.leiz...@huawei.com>; >>>> bhelg...@google.com; dw...@infradead.org; liubo (CU) >>>> <liub...@huawei.com>; r...@rjwysocki.net; robdcl...@gmail.com; >>>> hanjun@linaro.org; Sudeep Holla <sudeep.ho...@arm.com>; Robin >>>> Murphy <robin.mur...@arm.com>; nwatt...@codeaurora.org; Linuxarm >>>> <linux...@huawei.com> >>>> Subject: Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for >>>> Substream IDs >>>> >>>> Hi Shameer, >>>> >>>> On Thu, Nov 02, 2017 at 12:49:32PM +, Shameerali Kolothum Thodi wrote: >>>>> We had a go with this series on HiSIlicon D05 platform which doesn't have >>>>> support for ssids/ATS/PRI, to make sure it generally works. >>>>> >>>>> But observed the below crash on boot, >>>>> >>>>> [ 16.009084] WARNING: CPU: 59 PID: 391 at mm/page_alloc.c:3883 >>>> __alloc_pages_nodemask+0x19c/0xc48 >>>>> [ 16.026797] Modules linked in: >>>>> [ 16.032944] CPU: 59 PID: 391 Comm: kworker/59:1 Not tainted 4.14.0-rc1- >>>> 159539-ge42aca3 #236 >>>>> [...] >>>>> [ 16.068206] Workqueue: events deferred_probe_work_func >>>>> [ 16.078557] task: 8017d38a task.stack: 0b198000 >>>>> [ 16.090486] PC is at __alloc_pages_nodemask+0x19c/0xc48 >>>>> [ 16.101013] LR is at __alloc_pages_nodemask+0xe0/0xc48 >>>>> [ 16.469220] [] __alloc_pages_nodemask+0x19c/0xc48 >>>>> [ 16.481854] [] alloc_pages_current+0x80/0xcc >>>>> [ 16.493607] [] __get_free_pages+0xc/0x38 >>>>> [ 16.504661] [] swiotlb_alloc_coherent+0x64/0x190 >>>>> [ 16.517117] [] __dma_alloc+0x110/0x204 >>>>> [ 16.527820] [] dmam_alloc_coherent+0x88/0xf0 >>>>> [ 16.539575] [] >>>> arm_smmu_domain_finalise_s1+0x60/0x248 >>>>> [ 16.552909] [] arm_smmu_attach_dev+0x264/0x300 >>>>> [ 16.565013] [] __iommu_attach_device+0x48/0x5c >>>>> [ 16.577117] [] iommu_group_add_device+0x144/0x3a4 >>>>> [ 16.589746] [] iommu_group_get_for_dev+0x70/0xf8 >>>>> [ 16.602201] [] arm_smmu_add_device+0x1a4/0x418 >>>>> [ 16.614308] [] iort_iommu_configure+0xf0/0x16c >>>>> [ 16.626416] [] acpi_dma_configure+0x30/0x70 >>>>> [ 16.637994] [] dma_configure+0xa8/0xd4 >>>>> [ 16.648695] [] driver_probe_device+0x1a4/0x2dc >>>>> [ 16.673081] [] bus_for_each_drv+0x54/0x94 >>>>> [ 16.684307] [] __device_attach+0xc4/0x12c >>>>> [ 16.695533] [] device_initial_probe+0x10/0x18 >>>>> [ 16.707462] [] bus_probe_device+0x90/0x98 >>>>> >>>>> After a bit of debug it looks like on platforms where ssid is not >>>>> supported, >>>>> s1_cfg.num_contexts is set to zero and it eventually results in this crash >>>>> in, >>>>> arm_smmu_domain_finalise_s1() -->arm_smmu_alloc_cd_tables()--> >>>>> arm_smmu_alloc_cd_leaf_table() as num_leaf_entries is zero. >>>>> >>>>> With the below fix, it works on D05 now, >>>>> >>>>> diff --git a/drivers/i
RE: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for Substream IDs
> -Original Message- > From: Jean-Philippe Brucker [mailto:jean-philippe.bruc...@arm.com] > Sent: Friday, November 03, 2017 9:37 AM > To: xieyisheng (A) <xieyishe...@huawei.com>; Shameerali Kolothum Thodi > <shameerali.kolothum.th...@huawei.com> > Cc: linux-arm-ker...@lists.infradead.org; linux-...@vger.kernel.org; linux- > a...@vger.kernel.org; devicet...@vger.kernel.org; iommu@lists.linux- > foundation.org; Mark Rutland <mark.rutl...@arm.com>; Gabriele Paoloni > <gabriele.paol...@huawei.com>; Catalin Marinas > <catalin.mari...@arm.com>; Will Deacon <will.dea...@arm.com>; > ok...@codeaurora.org; yi.l@intel.com; Lorenzo Pieralisi > <lorenzo.pieral...@arm.com>; ashok@intel.com; t...@semihalf.com; > j...@8bytes.org; rfr...@cavium.com; l...@kernel.org; > jacob.jun@linux.intel.com; alex.william...@redhat.com; > robh...@kernel.org; Leizhen (ThunderTown) <thunder.leiz...@huawei.com>; > bhelg...@google.com; dw...@infradead.org; liubo (CU) > <liub...@huawei.com>; r...@rjwysocki.net; robdcl...@gmail.com; > hanjun@linaro.org; Sudeep Holla <sudeep.ho...@arm.com>; Robin > Murphy <robin.mur...@arm.com>; nwatt...@codeaurora.org; Linuxarm > <linux...@huawei.com> > Subject: Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for > Substream IDs > > On 03/11/17 05:45, Yisheng Xie wrote: > > Hi Jean, > > > > On 2017/11/3 1:02, Shameerali Kolothum Thodi wrote: > >> > >> > >>> -Original Message- > >>> From: Jean-Philippe Brucker [mailto:jean-philippe.bruc...@arm.com] > >>> Sent: Thursday, November 02, 2017 3:52 PM > >>> To: Shameerali Kolothum Thodi > <shameerali.kolothum.th...@huawei.com> > >>> Cc: linux-arm-ker...@lists.infradead.org; linux-...@vger.kernel.org; > >>> linux- > >>> a...@vger.kernel.org; devicet...@vger.kernel.org; iommu@lists.linux- > >>> foundation.org; Mark Rutland <mark.rutl...@arm.com>; xieyisheng (A) > >>> <xieyishe...@huawei.com>; Gabriele Paoloni > >>> <gabriele.paol...@huawei.com>; Catalin Marinas > >>> <catalin.mari...@arm.com>; Will Deacon <will.dea...@arm.com>; > >>> ok...@codeaurora.org; yi.l@intel.com; Lorenzo Pieralisi > >>> <lorenzo.pieral...@arm.com>; ashok@intel.com; t...@semihalf.com; > >>> j...@8bytes.org; rfr...@cavium.com; l...@kernel.org; > >>> jacob.jun@linux.intel.com; alex.william...@redhat.com; > >>> robh...@kernel.org; Leizhen (ThunderTown) > <thunder.leiz...@huawei.com>; > >>> bhelg...@google.com; dw...@infradead.org; liubo (CU) > >>> <liub...@huawei.com>; r...@rjwysocki.net; robdcl...@gmail.com; > >>> hanjun@linaro.org; Sudeep Holla <sudeep.ho...@arm.com>; Robin > >>> Murphy <robin.mur...@arm.com>; nwatt...@codeaurora.org; Linuxarm > >>> <linux...@huawei.com> > >>> Subject: Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for > >>> Substream IDs > >>> > >>> Hi Shameer, > >>> > >>> On Thu, Nov 02, 2017 at 12:49:32PM +, Shameerali Kolothum Thodi > wrote: > >>>> We had a go with this series on HiSIlicon D05 platform which doesn't have > >>>> support for ssids/ATS/PRI, to make sure it generally works. > >>>> > >>>> But observed the below crash on boot, > >>>> > >>>> [ 16.009084] WARNING: CPU: 59 PID: 391 at mm/page_alloc.c:3883 > >>> __alloc_pages_nodemask+0x19c/0xc48 > >>>> [ 16.026797] Modules linked in: > >>>> [ 16.032944] CPU: 59 PID: 391 Comm: kworker/59:1 Not tainted 4.14.0- > rc1- > >>> 159539-ge42aca3 #236 > >>>> [...] > >>>> [ 16.068206] Workqueue: events deferred_probe_work_func > >>>> [ 16.078557] task: 8017d38a task.stack: 0b198000 > >>>> [ 16.090486] PC is at __alloc_pages_nodemask+0x19c/0xc48 > >>>> [ 16.101013] LR is at __alloc_pages_nodemask+0xe0/0xc48 > >>>> [ 16.469220] [] > __alloc_pages_nodemask+0x19c/0xc48 > >>>> [ 16.481854] [] alloc_pages_current+0x80/0xcc > >>>> [ 16.493607] [] __get_free_pages+0xc/0x38 > >>>> [ 16.504661] [] swiotlb_alloc_coherent+0x64/0x190 > >>>> [ 16.517117] [] __dma_alloc+0x110/0x204 > >>>> [ 16.527820] [] dmam_alloc_coherent+0x88/0xf0 > >>>> [ 16.539575] [] > >>> arm_smmu_domain_finalise_s1
Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for Substream IDs
On 03/11/17 05:45, Yisheng Xie wrote: > Hi Jean, > > On 2017/11/3 1:02, Shameerali Kolothum Thodi wrote: >> >> >>> -Original Message- >>> From: Jean-Philippe Brucker [mailto:jean-philippe.bruc...@arm.com] >>> Sent: Thursday, November 02, 2017 3:52 PM >>> To: Shameerali Kolothum Thodi <shameerali.kolothum.th...@huawei.com> >>> Cc: linux-arm-ker...@lists.infradead.org; linux-...@vger.kernel.org; linux- >>> a...@vger.kernel.org; devicet...@vger.kernel.org; iommu@lists.linux- >>> foundation.org; Mark Rutland <mark.rutl...@arm.com>; xieyisheng (A) >>> <xieyishe...@huawei.com>; Gabriele Paoloni >>> <gabriele.paol...@huawei.com>; Catalin Marinas >>> <catalin.mari...@arm.com>; Will Deacon <will.dea...@arm.com>; >>> ok...@codeaurora.org; yi.l@intel.com; Lorenzo Pieralisi >>> <lorenzo.pieral...@arm.com>; ashok@intel.com; t...@semihalf.com; >>> j...@8bytes.org; rfr...@cavium.com; l...@kernel.org; >>> jacob.jun@linux.intel.com; alex.william...@redhat.com; >>> robh...@kernel.org; Leizhen (ThunderTown) <thunder.leiz...@huawei.com>; >>> bhelg...@google.com; dw...@infradead.org; liubo (CU) >>> <liub...@huawei.com>; r...@rjwysocki.net; robdcl...@gmail.com; >>> hanjun....@linaro.org; Sudeep Holla <sudeep.ho...@arm.com>; Robin >>> Murphy <robin.mur...@arm.com>; nwatt...@codeaurora.org; Linuxarm >>> <linux...@huawei.com> >>> Subject: Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for >>> Substream IDs >>> >>> Hi Shameer, >>> >>> On Thu, Nov 02, 2017 at 12:49:32PM +, Shameerali Kolothum Thodi wrote: >>>> We had a go with this series on HiSIlicon D05 platform which doesn't have >>>> support for ssids/ATS/PRI, to make sure it generally works. >>>> >>>> But observed the below crash on boot, >>>> >>>> [ 16.009084] WARNING: CPU: 59 PID: 391 at mm/page_alloc.c:3883 >>> __alloc_pages_nodemask+0x19c/0xc48 >>>> [ 16.026797] Modules linked in: >>>> [ 16.032944] CPU: 59 PID: 391 Comm: kworker/59:1 Not tainted 4.14.0-rc1- >>> 159539-ge42aca3 #236 >>>> [...] >>>> [ 16.068206] Workqueue: events deferred_probe_work_func >>>> [ 16.078557] task: 8017d38a task.stack: 0b198000 >>>> [ 16.090486] PC is at __alloc_pages_nodemask+0x19c/0xc48 >>>> [ 16.101013] LR is at __alloc_pages_nodemask+0xe0/0xc48 >>>> [ 16.469220] [] __alloc_pages_nodemask+0x19c/0xc48 >>>> [ 16.481854] [] alloc_pages_current+0x80/0xcc >>>> [ 16.493607] [] __get_free_pages+0xc/0x38 >>>> [ 16.504661] [] swiotlb_alloc_coherent+0x64/0x190 >>>> [ 16.517117] [] __dma_alloc+0x110/0x204 >>>> [ 16.527820] [] dmam_alloc_coherent+0x88/0xf0 >>>> [ 16.539575] [] >>> arm_smmu_domain_finalise_s1+0x60/0x248 >>>> [ 16.552909] [] arm_smmu_attach_dev+0x264/0x300 >>>> [ 16.565013] [] __iommu_attach_device+0x48/0x5c >>>> [ 16.577117] [] iommu_group_add_device+0x144/0x3a4 >>>> [ 16.589746] [] iommu_group_get_for_dev+0x70/0xf8 >>>> [ 16.602201] [] arm_smmu_add_device+0x1a4/0x418 >>>> [ 16.614308] [] iort_iommu_configure+0xf0/0x16c >>>> [ 16.626416] [] acpi_dma_configure+0x30/0x70 >>>> [ 16.637994] [] dma_configure+0xa8/0xd4 >>>> [ 16.648695] [] driver_probe_device+0x1a4/0x2dc >>>> [ 16.673081] [] bus_for_each_drv+0x54/0x94 >>>> [ 16.684307] [] __device_attach+0xc4/0x12c >>>> [ 16.695533] [] device_initial_probe+0x10/0x18 >>>> [ 16.707462] [] bus_probe_device+0x90/0x98 >>>> >>>> After a bit of debug it looks like on platforms where ssid is not >>>> supported, >>>> s1_cfg.num_contexts is set to zero and it eventually results in this crash >>>> in, >>>> arm_smmu_domain_finalise_s1() -->arm_smmu_alloc_cd_tables()--> >>>> arm_smmu_alloc_cd_leaf_table() as num_leaf_entries is zero. >>>> >>>> With the below fix, it works on D05 now, >>>> >>>> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c >>>> index 8ad90e2..51f5821 100644 >>>> --- a/drivers/iommu/arm-smmu-v3.c >>>> +++ b/drivers/iommu/arm-smmu-v3.c >>>> @@ -2433,7 +2433,10 @@ static int arm_smmu_domain_finalise(struct >>> iommu_domain *domain, >>>> domain->min_pa
Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for Substream IDs
Hi Jean, On 2017/11/3 1:02, Shameerali Kolothum Thodi wrote: > > >> -Original Message- >> From: Jean-Philippe Brucker [mailto:jean-philippe.bruc...@arm.com] >> Sent: Thursday, November 02, 2017 3:52 PM >> To: Shameerali Kolothum Thodi <shameerali.kolothum.th...@huawei.com> >> Cc: linux-arm-ker...@lists.infradead.org; linux-...@vger.kernel.org; linux- >> a...@vger.kernel.org; devicet...@vger.kernel.org; iommu@lists.linux- >> foundation.org; Mark Rutland <mark.rutl...@arm.com>; xieyisheng (A) >> <xieyishe...@huawei.com>; Gabriele Paoloni >> <gabriele.paol...@huawei.com>; Catalin Marinas >> <catalin.mari...@arm.com>; Will Deacon <will.dea...@arm.com>; >> ok...@codeaurora.org; yi.l@intel.com; Lorenzo Pieralisi >> <lorenzo.pieral...@arm.com>; ashok@intel.com; t...@semihalf.com; >> j...@8bytes.org; rfr...@cavium.com; l...@kernel.org; >> jacob.jun@linux.intel.com; alex.william...@redhat.com; >> robh...@kernel.org; Leizhen (ThunderTown) <thunder.leiz...@huawei.com>; >> bhelg...@google.com; dw...@infradead.org; liubo (CU) >> <liub...@huawei.com>; r...@rjwysocki.net; robdcl...@gmail.com; >> hanjun....@linaro.org; Sudeep Holla <sudeep.ho...@arm.com>; Robin >> Murphy <robin.mur...@arm.com>; nwatt...@codeaurora.org; Linuxarm >> <linux...@huawei.com> >> Subject: Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for >> Substream IDs >> >> Hi Shameer, >> >> On Thu, Nov 02, 2017 at 12:49:32PM +, Shameerali Kolothum Thodi wrote: >>> We had a go with this series on HiSIlicon D05 platform which doesn't have >>> support for ssids/ATS/PRI, to make sure it generally works. >>> >>> But observed the below crash on boot, >>> >>> [ 16.009084] WARNING: CPU: 59 PID: 391 at mm/page_alloc.c:3883 >> __alloc_pages_nodemask+0x19c/0xc48 >>> [ 16.026797] Modules linked in: >>> [ 16.032944] CPU: 59 PID: 391 Comm: kworker/59:1 Not tainted 4.14.0-rc1- >> 159539-ge42aca3 #236 >>> [...] >>> [ 16.068206] Workqueue: events deferred_probe_work_func >>> [ 16.078557] task: 8017d38a task.stack: 0b198000 >>> [ 16.090486] PC is at __alloc_pages_nodemask+0x19c/0xc48 >>> [ 16.101013] LR is at __alloc_pages_nodemask+0xe0/0xc48 >>> [ 16.469220] [] __alloc_pages_nodemask+0x19c/0xc48 >>> [ 16.481854] [] alloc_pages_current+0x80/0xcc >>> [ 16.493607] [] __get_free_pages+0xc/0x38 >>> [ 16.504661] [] swiotlb_alloc_coherent+0x64/0x190 >>> [ 16.517117] [] __dma_alloc+0x110/0x204 >>> [ 16.527820] [] dmam_alloc_coherent+0x88/0xf0 >>> [ 16.539575] [] >> arm_smmu_domain_finalise_s1+0x60/0x248 >>> [ 16.552909] [] arm_smmu_attach_dev+0x264/0x300 >>> [ 16.565013] [] __iommu_attach_device+0x48/0x5c >>> [ 16.577117] [] iommu_group_add_device+0x144/0x3a4 >>> [ 16.589746] [] iommu_group_get_for_dev+0x70/0xf8 >>> [ 16.602201] [] arm_smmu_add_device+0x1a4/0x418 >>> [ 16.614308] [] iort_iommu_configure+0xf0/0x16c >>> [ 16.626416] [] acpi_dma_configure+0x30/0x70 >>> [ 16.637994] [] dma_configure+0xa8/0xd4 >>> [ 16.648695] [] driver_probe_device+0x1a4/0x2dc >>> [ 16.673081] [] bus_for_each_drv+0x54/0x94 >>> [ 16.684307] [] __device_attach+0xc4/0x12c >>> [ 16.695533] [] device_initial_probe+0x10/0x18 >>> [ 16.707462] [] bus_probe_device+0x90/0x98 >>> >>> After a bit of debug it looks like on platforms where ssid is not supported, >>> s1_cfg.num_contexts is set to zero and it eventually results in this crash >>> in, >>> arm_smmu_domain_finalise_s1() -->arm_smmu_alloc_cd_tables()--> >>> arm_smmu_alloc_cd_leaf_table() as num_leaf_entries is zero. >>> >>> With the below fix, it works on D05 now, >>> >>> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c >>> index 8ad90e2..51f5821 100644 >>> --- a/drivers/iommu/arm-smmu-v3.c >>> +++ b/drivers/iommu/arm-smmu-v3.c >>> @@ -2433,7 +2433,10 @@ static int arm_smmu_domain_finalise(struct >> iommu_domain *domain, >>> domain->min_pasid = 1; >>> domain->max_pasid = master->num_ssids - 1; >>> smmu_domain->s1_cfg.num_contexts = >>> master->num_ssids; >>> + } else { >>> + smmu_domain->s1_cfg.num_contexts = 1; >>> } >>> + >>>
RE: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for Substream IDs
> -Original Message- > From: Jean-Philippe Brucker [mailto:jean-philippe.bruc...@arm.com] > Sent: Thursday, November 02, 2017 3:52 PM > To: Shameerali Kolothum Thodi <shameerali.kolothum.th...@huawei.com> > Cc: linux-arm-ker...@lists.infradead.org; linux-...@vger.kernel.org; linux- > a...@vger.kernel.org; devicet...@vger.kernel.org; iommu@lists.linux- > foundation.org; Mark Rutland <mark.rutl...@arm.com>; xieyisheng (A) > <xieyishe...@huawei.com>; Gabriele Paoloni > <gabriele.paol...@huawei.com>; Catalin Marinas > <catalin.mari...@arm.com>; Will Deacon <will.dea...@arm.com>; > ok...@codeaurora.org; yi.l@intel.com; Lorenzo Pieralisi > <lorenzo.pieral...@arm.com>; ashok@intel.com; t...@semihalf.com; > j...@8bytes.org; rfr...@cavium.com; l...@kernel.org; > jacob.jun@linux.intel.com; alex.william...@redhat.com; > robh...@kernel.org; Leizhen (ThunderTown) <thunder.leiz...@huawei.com>; > bhelg...@google.com; dw...@infradead.org; liubo (CU) > <liub...@huawei.com>; r...@rjwysocki.net; robdcl...@gmail.com; > hanjun@linaro.org; Sudeep Holla <sudeep.ho...@arm.com>; Robin > Murphy <robin.mur...@arm.com>; nwatt...@codeaurora.org; Linuxarm > <linux...@huawei.com> > Subject: Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for > Substream IDs > > Hi Shameer, > > On Thu, Nov 02, 2017 at 12:49:32PM +, Shameerali Kolothum Thodi wrote: > > We had a go with this series on HiSIlicon D05 platform which doesn't have > > support for ssids/ATS/PRI, to make sure it generally works. > > > > But observed the below crash on boot, > > > > [ 16.009084] WARNING: CPU: 59 PID: 391 at mm/page_alloc.c:3883 > __alloc_pages_nodemask+0x19c/0xc48 > > [ 16.026797] Modules linked in: > > [ 16.032944] CPU: 59 PID: 391 Comm: kworker/59:1 Not tainted 4.14.0-rc1- > 159539-ge42aca3 #236 > > [...] > > [ 16.068206] Workqueue: events deferred_probe_work_func > > [ 16.078557] task: 8017d38a task.stack: 0b198000 > > [ 16.090486] PC is at __alloc_pages_nodemask+0x19c/0xc48 > > [ 16.101013] LR is at __alloc_pages_nodemask+0xe0/0xc48 > > [ 16.469220] [] __alloc_pages_nodemask+0x19c/0xc48 > > [ 16.481854] [] alloc_pages_current+0x80/0xcc > > [ 16.493607] [] __get_free_pages+0xc/0x38 > > [ 16.504661] [] swiotlb_alloc_coherent+0x64/0x190 > > [ 16.517117] [] __dma_alloc+0x110/0x204 > > [ 16.527820] [] dmam_alloc_coherent+0x88/0xf0 > > [ 16.539575] [] > arm_smmu_domain_finalise_s1+0x60/0x248 > > [ 16.552909] [] arm_smmu_attach_dev+0x264/0x300 > > [ 16.565013] [] __iommu_attach_device+0x48/0x5c > > [ 16.577117] [] iommu_group_add_device+0x144/0x3a4 > > [ 16.589746] [] iommu_group_get_for_dev+0x70/0xf8 > > [ 16.602201] [] arm_smmu_add_device+0x1a4/0x418 > > [ 16.614308] [] iort_iommu_configure+0xf0/0x16c > > [ 16.626416] [] acpi_dma_configure+0x30/0x70 > > [ 16.637994] [] dma_configure+0xa8/0xd4 > > [ 16.648695] [] driver_probe_device+0x1a4/0x2dc > > [ 16.673081] [] bus_for_each_drv+0x54/0x94 > > [ 16.684307] [] __device_attach+0xc4/0x12c > > [ 16.695533] [] device_initial_probe+0x10/0x18 > > [ 16.707462] [] bus_probe_device+0x90/0x98 > > > > After a bit of debug it looks like on platforms where ssid is not supported, > > s1_cfg.num_contexts is set to zero and it eventually results in this crash > > in, > > arm_smmu_domain_finalise_s1() -->arm_smmu_alloc_cd_tables()--> > > arm_smmu_alloc_cd_leaf_table() as num_leaf_entries is zero. > > > > With the below fix, it works on D05 now, > > > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > > index 8ad90e2..51f5821 100644 > > --- a/drivers/iommu/arm-smmu-v3.c > > +++ b/drivers/iommu/arm-smmu-v3.c > > @@ -2433,7 +2433,10 @@ static int arm_smmu_domain_finalise(struct > iommu_domain *domain, > > domain->min_pasid = 1; > > domain->max_pasid = master->num_ssids - 1; > > smmu_domain->s1_cfg.num_contexts = > > master->num_ssids; > > + } else { > > + smmu_domain->s1_cfg.num_contexts = 1; > > } > > + > > smmu_domain->s1_cfg.can_stall = master->ste.can_stall; > > break; > > case ARM_SMMU_DOMAIN_NESTED: > > > > > > I am not sure this is right place do this. Please take a look. > > Thanks for testing the series and reporting the bug. I added the > following patch to branch svm/curre
Re: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for Substream IDs
Hi Shameer, On Thu, Nov 02, 2017 at 12:49:32PM +, Shameerali Kolothum Thodi wrote: > We had a go with this series on HiSIlicon D05 platform which doesn't have > support for ssids/ATS/PRI, to make sure it generally works. > > But observed the below crash on boot, > > [ 16.009084] WARNING: CPU: 59 PID: 391 at mm/page_alloc.c:3883 > __alloc_pages_nodemask+0x19c/0xc48 > [ 16.026797] Modules linked in: > [ 16.032944] CPU: 59 PID: 391 Comm: kworker/59:1 Not tainted > 4.14.0-rc1-159539-ge42aca3 #236 > [...] > [ 16.068206] Workqueue: events deferred_probe_work_func > [ 16.078557] task: 8017d38a task.stack: 0b198000 > [ 16.090486] PC is at __alloc_pages_nodemask+0x19c/0xc48 > [ 16.101013] LR is at __alloc_pages_nodemask+0xe0/0xc48 > [ 16.469220] [] __alloc_pages_nodemask+0x19c/0xc48 > [ 16.481854] [] alloc_pages_current+0x80/0xcc > [ 16.493607] [] __get_free_pages+0xc/0x38 > [ 16.504661] [] swiotlb_alloc_coherent+0x64/0x190 > [ 16.517117] [] __dma_alloc+0x110/0x204 > [ 16.527820] [] dmam_alloc_coherent+0x88/0xf0 > [ 16.539575] [] arm_smmu_domain_finalise_s1+0x60/0x248 > [ 16.552909] [] arm_smmu_attach_dev+0x264/0x300 > [ 16.565013] [] __iommu_attach_device+0x48/0x5c > [ 16.577117] [] iommu_group_add_device+0x144/0x3a4 > [ 16.589746] [] iommu_group_get_for_dev+0x70/0xf8 > [ 16.602201] [] arm_smmu_add_device+0x1a4/0x418 > [ 16.614308] [] iort_iommu_configure+0xf0/0x16c > [ 16.626416] [] acpi_dma_configure+0x30/0x70 > [ 16.637994] [] dma_configure+0xa8/0xd4 > [ 16.648695] [] driver_probe_device+0x1a4/0x2dc > [ 16.673081] [] bus_for_each_drv+0x54/0x94 > [ 16.684307] [] __device_attach+0xc4/0x12c > [ 16.695533] [] device_initial_probe+0x10/0x18 > [ 16.707462] [] bus_probe_device+0x90/0x98 > > After a bit of debug it looks like on platforms where ssid is not supported, > s1_cfg.num_contexts is set to zero and it eventually results in this crash > in, > arm_smmu_domain_finalise_s1() -->arm_smmu_alloc_cd_tables()--> > arm_smmu_alloc_cd_leaf_table() as num_leaf_entries is zero. > > With the below fix, it works on D05 now, > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > index 8ad90e2..51f5821 100644 > --- a/drivers/iommu/arm-smmu-v3.c > +++ b/drivers/iommu/arm-smmu-v3.c > @@ -2433,7 +2433,10 @@ static int arm_smmu_domain_finalise(struct > iommu_domain *domain, > domain->min_pasid = 1; > domain->max_pasid = master->num_ssids - 1; > smmu_domain->s1_cfg.num_contexts = master->num_ssids; > + } else { > + smmu_domain->s1_cfg.num_contexts = 1; > } > + > smmu_domain->s1_cfg.can_stall = master->ste.can_stall; > break; > case ARM_SMMU_DOMAIN_NESTED: > > > I am not sure this is right place do this. Please take a look. Thanks for testing the series and reporting the bug. I added the following patch to branch svm/current, does it work for you? Thanks, Jean diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c index 42c8378624ed..edda466adc81 100644 --- a/drivers/iommu/arm-smmu-v3.c +++ b/drivers/iommu/arm-smmu-v3.c @@ -3169,9 +3169,7 @@ static int arm_smmu_add_device(struct device *dev) } } - if (smmu->ssid_bits) - master->num_ssids = 1 << min(smmu->ssid_bits, -fwspec->num_pasid_bits); + master->num_ssids = 1 << min(smmu->ssid_bits, fwspec->num_pasid_bits); if (fwspec->can_stall && smmu->features & ARM_SMMU_FEAT_STALLS) { master->can_fault = true; ___ iommu mailing list iommu@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/iommu
RE: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for Substream IDs
Hi Jean, > -Original Message- > From: linux-arm-kernel [mailto:linux-arm-kernel-boun...@lists.infradead.org] > On Behalf Of Jean-Philippe Brucker > Sent: Friday, October 06, 2017 2:32 PM > To: linux-arm-ker...@lists.infradead.org; linux-...@vger.kernel.org; linux- > a...@vger.kernel.org; devicet...@vger.kernel.org; iommu@lists.linux- > foundation.org > Cc: mark.rutl...@arm.com; xieyisheng (A); > Gabriele Paoloni ; catalin.mari...@arm.com; > will.dea...@arm.com; ok...@codeaurora.org; yi.l@intel.com; > lorenzo.pieral...@arm.com; ashok@intel.com; t...@semihalf.com; > j...@8bytes.org; rfr...@cavium.com; l...@kernel.org; > jacob.jun@linux.intel.com; alex.william...@redhat.com; > robh...@kernel.org; Leizhen (ThunderTown) ; > bhelg...@google.com; dw...@infradead.org; liubo (CU) > ; r...@rjwysocki.net; robdcl...@gmail.com; > hanjun@linaro.org; sudeep.ho...@arm.com; robin.mur...@arm.com; > nwatt...@codeaurora.org > Subject: [RFCv2 PATCH 14/36] iommu/arm-smmu-v3: Add support for > Substream IDs > > At the moment, the SMMUv3 driver offers only one stage-1 or stage-2 > address space to each device. SMMUv3 allows to associate multiple address > spaces per device. In addition to the Stream ID (SID), that identifies a > device, we can now have Substream IDs (SSID) identifying an address space. > In PCIe lingo, SID is called Requester ID (RID) and SSID is called Process > Address-Space ID (PASID). We had a go with this series on HiSIlicon D05 platform which doesn't have support for ssids/ATS/PRI, to make sure it generally works. But observed the below crash on boot, [ 16.009084] WARNING: CPU: 59 PID: 391 at mm/page_alloc.c:3883 __alloc_pages_nodemask+0x19c/0xc48 [ 16.026797] Modules linked in: [ 16.032944] CPU: 59 PID: 391 Comm: kworker/59:1 Not tainted 4.14.0-rc1-159539-ge42aca3 #236 [...] [ 16.068206] Workqueue: events deferred_probe_work_func [ 16.078557] task: 8017d38a task.stack: 0b198000 [ 16.090486] PC is at __alloc_pages_nodemask+0x19c/0xc48 [ 16.101013] LR is at __alloc_pages_nodemask+0xe0/0xc48 [ 16.469220] [] __alloc_pages_nodemask+0x19c/0xc48 [ 16.481854] [] alloc_pages_current+0x80/0xcc [ 16.493607] [] __get_free_pages+0xc/0x38 [ 16.504661] [] swiotlb_alloc_coherent+0x64/0x190 [ 16.517117] [] __dma_alloc+0x110/0x204 [ 16.527820] [] dmam_alloc_coherent+0x88/0xf0 [ 16.539575] [] arm_smmu_domain_finalise_s1+0x60/0x248 [ 16.552909] [] arm_smmu_attach_dev+0x264/0x300 [ 16.565013] [] __iommu_attach_device+0x48/0x5c [ 16.577117] [] iommu_group_add_device+0x144/0x3a4 [ 16.589746] [] iommu_group_get_for_dev+0x70/0xf8 [ 16.602201] [] arm_smmu_add_device+0x1a4/0x418 [ 16.614308] [] iort_iommu_configure+0xf0/0x16c [ 16.626416] [] acpi_dma_configure+0x30/0x70 [ 16.637994] [] dma_configure+0xa8/0xd4 [ 16.648695] [] driver_probe_device+0x1a4/0x2dc [ 16.673081] [] bus_for_each_drv+0x54/0x94 [ 16.684307] [] __device_attach+0xc4/0x12c [ 16.695533] [] device_initial_probe+0x10/0x18 [ 16.707462] [] bus_probe_device+0x90/0x98 After a bit of debug it looks like on platforms where ssid is not supported, s1_cfg.num_contexts is set to zero and it eventually results in this crash in, arm_smmu_domain_finalise_s1() -->arm_smmu_alloc_cd_tables()--> arm_smmu_alloc_cd_leaf_table() as num_leaf_entries is zero. With the below fix, it works on D05 now, diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c index 8ad90e2..51f5821 100644 --- a/drivers/iommu/arm-smmu-v3.c +++ b/drivers/iommu/arm-smmu-v3.c @@ -2433,7 +2433,10 @@ static int arm_smmu_domain_finalise(struct iommu_domain *domain, domain->min_pasid = 1; domain->max_pasid = master->num_ssids - 1; smmu_domain->s1_cfg.num_contexts = master->num_ssids; + } else { + smmu_domain->s1_cfg.num_contexts = 1; } + smmu_domain->s1_cfg.can_stall = master->ste.can_stall; break; case ARM_SMMU_DOMAIN_NESTED: I am not sure this is right place do this. Please take a look. Thanks, Shameer ___ iommu mailing list iommu@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/iommu