iommu/vt-d: Fixes for v6.1-rc2
Subsystems (11)
Lu Baolu
maintainerHi Joerg,
Below fixes are queued for v6.1. They aim to fix:
- A lockdep splat issue in intel_iommu_init().
- Allow NVS regions to pass RMRR check.
- Domain cleanup in error path.
This series is also available at github.
https://github.com/LuBaolu/intel-iommu/commits/vtd-fix-for-v6.1-rc2
Please consider it for the iommu/fix branch.
Best regards,
Lu Baolu
Charlotte Tan (1):
iommu/vt-d: Allow NVS regions in arch_rmrr_sanity_check()
Jerry Snitselaar (1):
iommu/vt-d: Clean up si_domain in the init_dmars() error path
Lu Baolu (2):
iommu: Add gfp parameter to iommu_alloc_resv_region iommu/vt-d: Use rcu_lock in get_resv_regions
include/linux/iommu.h | 2 +- arch/x86/include/asm/iommu.h | 4 +++- drivers/acpi/arm64/iort.c | 3 ++- drivers/iommu/amd/iommu.c | 7 ++++--- drivers/iommu/apple-dart.c | 2 +- drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 2 +- drivers/iommu/arm/arm-smmu/arm-smmu.c | 2 +- drivers/iommu/intel/iommu.c | 17 ++++++++++++----- drivers/iommu/iommu.c | 7 ++++--- drivers/iommu/mtk_iommu.c | 3 ++- drivers/iommu/virtio-iommu.c | 9 ++++++--- 11 files changed, 37 insertions(+), 21 deletions(-)
Signature · 1 line
2.34.1
Lu Baolu
maintainerPATCH 1/4 iommu: Add gfp parameter to iommu_alloc_resv_region
Add gfp parameter to iommu_alloc_resv_region() for the callers to specify the memory allocation behavior. Thus iommu_alloc_resv_region() could also be available in critical contexts.
--- include/linux/iommu.h | 2 +- drivers/acpi/arm64/iort.c | 3 ++- drivers/iommu/amd/iommu.c | 7 ++++--- drivers/iommu/apple-dart.c | 2 +- drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 2 +- drivers/iommu/arm/arm-smmu/arm-smmu.c | 2 +- drivers/iommu/intel/iommu.c | 8 +++++--- drivers/iommu/iommu.c | 7 ++++--- drivers/iommu/mtk_iommu.c | 3 ++- drivers/iommu/virtio-iommu.c | 9 ++++++--- 10 files changed, 27 insertions(+), 18 deletions(-)
include/linux/iommu.h +1 -1
@@ -455,7 +455,7 @@ extern void iommu_set_default_translated(bool cmd_line);extern bool iommu_default_passthrough (void );extern struct iommu_resv_region *iommu_alloc_resv_region (phys_addr_t start,size_t length,int prot, -enum iommu_resv_type type); +enum iommu_resv_type type,gfp_t gfp);extern int iommu_get_group_resv_regions (struct iommu_group *group,struct list_head *head);
drivers/acpi/arm64/iort.c +2 -1
@@ -1142,7 +1142,8 @@ static void iort_iommu_msi_get_resv_regions(struct device *dev,struct iommu_resv_region *region; region =iommu_alloc_resv_region (base +SZ_64K ,SZ_64K , - prot,IOMMU_RESV_MSI ); + prot,IOMMU_RESV_MSI , +GFP_KERNEL );if (region)list_add_tail (®ion->list, head); }
drivers/iommu/amd/iommu.c +4 -3
@@ -2330,7 +2330,8 @@ static void amd_iommu_get_resv_regions(struct device *dev, type =IOMMU_RESV_RESERVED ; region =iommu_alloc_resv_region (entry->address_start, - length, prot, type); + length, prot, type, +GFP_KERNEL );if (!region) {dev_err (dev,"Out of memory allocating dm-regions\n" );return ;
@@ -2340,14 +2341,14 @@ static void amd_iommu_get_resv_regions(struct device *dev, region =iommu_alloc_resv_region (MSI_RANGE_START ,MSI_RANGE_END -MSI_RANGE_START +1 , -0 ,IOMMU_RESV_MSI ); +0 ,IOMMU_RESV_MSI ,GFP_KERNEL );if (!region)return ;list_add_tail (®ion->list, head); region =iommu_alloc_resv_region (HT_RANGE_START ,HT_RANGE_END -HT_RANGE_START +1 , -0 ,IOMMU_RESV_RESERVED ); +0 ,IOMMU_RESV_RESERVED ,GFP_KERNEL );if (!region)return ;list_add_tail (®ion->list, head);
drivers/iommu/apple-dart.c +1 -1
@@ -758,7 +758,7 @@ static void apple_dart_get_resv_regions(struct device *dev, region =iommu_alloc_resv_region (DOORBELL_ADDR ,PAGE_SIZE , prot, -IOMMU_RESV_MSI ); +IOMMU_RESV_MSI ,GFP_KERNEL );if (!region)return ;
drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c +1 -1
@@ -2757,7 +2757,7 @@ static void arm_smmu_get_resv_regions(struct device *dev,int prot =IOMMU_WRITE |IOMMU_NOEXEC |IOMMU_MMIO ; region =iommu_alloc_resv_region (MSI_IOVA_BASE ,MSI_IOVA_LENGTH , - prot,IOMMU_RESV_SW_MSI ); + prot,IOMMU_RESV_SW_MSI ,GFP_KERNEL );if (!region)return ;
drivers/iommu/arm/arm-smmu/arm-smmu.c +1 -1
@@ -1534,7 +1534,7 @@ static void arm_smmu_get_resv_regions(struct device *dev,int prot =IOMMU_WRITE |IOMMU_NOEXEC |IOMMU_MMIO ; region =iommu_alloc_resv_region (MSI_IOVA_BASE ,MSI_IOVA_LENGTH , - prot,IOMMU_RESV_SW_MSI ); + prot,IOMMU_RESV_SW_MSI ,GFP_KERNEL );if (!region)return ;
drivers/iommu/intel/iommu.c +5 -3
@@ -4552,7 +4552,8 @@ static void intel_iommu_get_resv_regions(struct device *device, IOMMU_RESV_DIRECT_RELAXABLE :IOMMU_RESV_DIRECT ; resv =iommu_alloc_resv_region (rmrr->base_address, - length, prot, type); + length, prot, type, +GFP_KERNEL );if (!resv)break ;
@@ -4567,7 +4568,8 @@ static void intel_iommu_get_resv_regions(struct device *device,if ((pdev->class >>8 ) ==PCI_CLASS_BRIDGE_ISA ) { reg =iommu_alloc_resv_region (0 ,1UL <<24 , prot, -IOMMU_RESV_DIRECT_RELAXABLE ); +IOMMU_RESV_DIRECT_RELAXABLE , +GFP_KERNEL );if (reg)list_add_tail (®->list, head); }
@@ -4576,7 +4578,7 @@ static void intel_iommu_get_resv_regions(struct device *device, reg =iommu_alloc_resv_region (IOAPIC_RANGE_START ,IOAPIC_RANGE_END -IOAPIC_RANGE_START +1 , -0 ,IOMMU_RESV_MSI ); +0 ,IOMMU_RESV_MSI ,GFP_KERNEL );if (!reg)return ;list_add_tail (®->list, head);
drivers/iommu/iommu.c +4 -3
@@ -504,7 +504,7 @@ static int iommu_insert_resv_region(struct iommu_resv_region *new,LIST_HEAD (stack); nr =iommu_alloc_resv_region (new->start, new->length, - new->prot, new->type); + new->prot, new->type,GFP_KERNEL );if (!nr)return -ENOMEM ;
@@ -2579,11 +2579,12 @@ EXPORT_SYMBOL(iommu_put_resv_regions);struct iommu_resv_region *iommu_alloc_resv_region (phys_addr_t start,size_t length,int prot, -enum iommu_resv_type type) +enum iommu_resv_type type, +gfp_t gfp) {struct iommu_resv_region *region; - region =kzalloc (sizeof (*region),GFP_KERNEL ); + region =kzalloc (sizeof (*region), gfp);if (!region)return NULL ;
drivers/iommu/mtk_iommu.c +2 -1
@@ -917,7 +917,8 @@ static void mtk_iommu_get_resv_regions(struct device *dev,continue ; region =iommu_alloc_resv_region (resv->iova_base, resv->size, - prot,IOMMU_RESV_RESERVED ); + prot,IOMMU_RESV_RESERVED , +GFP_KERNEL );if (!region)return ;
drivers/iommu/virtio-iommu.c +6 -3
@@ -490,11 +490,13 @@ static int viommu_add_resv_mem(struct viommu_endpoint *vdev, fallthrough;case VIRTIO_IOMMU_RESV_MEM_T_RESERVED : region =iommu_alloc_resv_region (start, size,0 , -IOMMU_RESV_RESERVED ); +IOMMU_RESV_RESERVED , +GFP_KERNEL );break ;case VIRTIO_IOMMU_RESV_MEM_T_MSI : region =iommu_alloc_resv_region (start, size, prot, -IOMMU_RESV_MSI ); +IOMMU_RESV_MSI , +GFP_KERNEL );break ; }if (!region)
@@ -909,7 +911,8 @@ static void viommu_get_resv_regions(struct device *dev, struct list_head *head)*/ if (!msi) { msi =iommu_alloc_resv_region (MSI_IOVA_BASE ,MSI_IOVA_LENGTH , - prot,IOMMU_RESV_SW_MSI ); + prot,IOMMU_RESV_SW_MSI , +GFP_KERNEL );if (!msi)return ;
Signature · 1 line
2.34.1
Lu Baolu
maintainerPATCH 2/4 iommu/vt-d: Use rcu_lock in get_resv_regions
In reply to Lu BaoluCommit 5f64ce5411b46 ("iommu/vt-d: Duplicate iommu_resv_region objects per device list") converted rcu_lock in get_resv_regions to dmar_global_lock to allow sleeping in iommu_alloc_resv_region(). This introduced possible recursive locking if get_resv_regions is called from within a section where intel_iommu_init() already holds dmar_global_lock.
Especially, after commit 57365a04c921 ("iommu: Move bus setup to IOMMU device registration"), below lockdep splats could always be seen.
============================================
WARNING: possible recursive locking detected
6.0.0-rc4+ #325 Tainted: G I
--------------------------------------------
swapper/0/1 is trying to acquire lock:
ffffffffa8a18c90 (dmar_global_lock){++++}-{3:3}, at:
intel_iommu_get_resv_regions+0x25/0x270
but task is already holding lock:
ffffffffa8a18c90 (dmar_global_lock){++++}-{3:3}, at:
intel_iommu_init+0x36d/0x6ea
...
Call Trace: <TASK> dump_stack_lvl+0x48/0x5f __lock_acquire.cold.73+0xad/0x2bb lock_acquire+0xc2/0x2e0 ? intel_iommu_get_resv_regions+0x25/0x270 ? lock_is_held_type+0x9d/0x110 down_read+0x42/0x150 ? intel_iommu_get_resv_regions+0x25/0x270 intel_iommu_get_resv_regions+0x25/0x270 iommu_create_device_direct_mappings.isra.28+0x8d/0x1c0 ? iommu_get_dma_cookie+0x6d/0x90 bus_iommu_probe+0x19f/0x2e0 iommu_device_register+0xd4/0x130 intel_iommu_init+0x3e1/0x6ea ? iommu_setup+0x289/0x289 ? rdinit_setup+0x34/0x34 pci_iommu_init+0x12/0x3a do_one_initcall+0x65/0x320 ? rdinit_setup+0x34/0x34 ? rcu_read_lock_sched_held+0x5a/0x80 kernel_init_freeable+0x28a/0x2f3 ? rest_init+0x1b0/0x1b0 kernel_init+0x1a/0x130 ret_from_fork+0x1f/0x30 </TASK>
This rolls back dmar_global_lock to rcu_lock in get_resv_regions to avoid the lockdep splat.
--- drivers/iommu/intel/iommu.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/iommu/intel/iommu.c +3 -3
@@ -4534,7 +4534,7 @@ static void intel_iommu_get_resv_regions(struct device *device,struct device *i_dev;int i; -down_read (&dmar_global_lock); +rcu_read_lock ();for_each_rmrr_units (rmrr) {for_each_active_dev_scope (rmrr->devices, rmrr->devices_cnt, i, i_dev) {
@@ -4553,14 +4553,14 @@ static void intel_iommu_get_resv_regions(struct device *device, resv =iommu_alloc_resv_region (rmrr->base_address, length, prot, type, -GFP_KERNEL ); +GFP_ATOMIC );if (!resv)break ;list_add_tail (&resv->list, head); } } -up_read (&dmar_global_lock); +rcu_read_unlock ();#ifdef CONFIG_INTEL_IOMMU_FLOPPY_WA if (dev_is_pci (device)) {
Signature · 1 line
2.34.1
Lu Baolu
maintainerPATCH 3/4 iommu/vt-d: Allow NVS regions in arch_rmrr_sanity_check()
In reply to Lu BaoluFrom: Charlotte Tan <[email protected]>
arch_rmrr_sanity_check() warns if the RMRR is not covered by an ACPI Reserved region, but it seems like it should accept an NVS region as well. The ACPI spec
https://uefi.org/specs/ACPI/6.5/15_System_Address_Map_Interfaces.html
uses similar wording for "Reserved" and "NVS" region types; for NVS regions it says "This range of addresses is in use or reserved by the system and must not be used by the operating system."
There is an old comment on this mailing list that also suggests NVS regions should pass the arch_rmrr_sanity_check() test:
The warnings come from arch_rmrr_sanity_check() since it checks whether the region is E820_TYPE_RESERVED. However, if the purpose of the check is to detect RMRR has regions that may be used by OS as free memory, isn't E820_TYPE_NVS safe, too?
This patch overlaps with another proposed patch that would add the region type to the log since sometimes the bug reporter sees this log on the console but doesn't know to include the kernel log:
https://lore.kernel.org/lkml/[email protected]/
Here's an example of the "Firmware Bug" apparent false positive (wrapped for line length):
DMAR: [Firmware Bug]: No firmware reserved region can cover this RMRR
[0x000000006f760000-0x000000006f762fff], contact BIOS vendor for
fixes
DMAR: [Firmware Bug]: Your BIOS is broken; bad RMRR
[0x000000006f760000-0x000000006f762fff]This is the snippet from the e820 table:
BIOS-e820: [mem 0x0000000068bff000-0x000000006ebfefff] reserved
BIOS-e820: [mem 0x000000006ebff000-0x000000006f9fefff] ACPI NVS
BIOS-e820: [mem 0x000000006f9ff000-0x000000006fffefff] ACPI data
--- arch/x86/include/asm/iommu.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
arch/x86/include/asm/iommu.h +3 -1
@@ -25,8 +25,10 @@ arch_rmrr_sanity_check(struct acpi_dmar_reserved_memory *rmrr) {u64 start = rmrr->base_address;u64 end = rmrr->end_address +1 ; +int entry_type; -if (e820__mapped_all (start, end,E820_TYPE_RESERVED )) + entry_type =e820__get_entry_type (start, end); +if (entry_type ==E820_TYPE_RESERVED || entry_type ==E820_TYPE_NVS )return 0 ; pr_err(FW_BUG "No firmware reserved region can cover this RMRR [%#018Lx-%#018Lx], contact BIOS vendor for fixes\n",
Signature · 1 line
2.34.1
Lu Baolu
maintainerPATCH 4/4 iommu/vt-d: Clean up si_domain in the init_dmars() error path
In reply to Lu BaoluFrom: Jerry Snitselaar <[email protected]>
A splat from kmem_cache_destroy() was seen with a kernel prior to commit ee2653bbe89d ("iommu/vt-d: Remove domain and devinfo mempool") when there was a failure in init_dmars(), because the iommu_domain cache still had objects. While the mempool code is now gone, there still is a leak of the si_domain memory if init_dmars() fails. So clean up si_domain in the init_dmars() error path.
--- drivers/iommu/intel/iommu.c | 5 +++++ 1 file changed, 5 insertions(+)
drivers/iommu/intel/iommu.c +5 -0
@@ -2410,6 +2410,7 @@ static int __init si_domain_init(int hw)if (md_domain_init (si_domain,DEFAULT_DOMAIN_ADDRESS_WIDTH )) {domain_exit (si_domain); + si_domain =NULL ;return -EFAULT ; }
@@ -3052,6 +3053,10 @@ static int __init init_dmars(void)disable_dmar_iommu (iommu);free_dmar_iommu (iommu); } +if (si_domain) { +domain_exit (si_domain); + si_domain =NULL ; + }return ret; }
Signature · 1 line
2.34.1
Quoted 10 lines · Lu Baolu
Hi Joerg,
Below fixes are queued for v6.1. They aim to fix:
- A lockdep splat issue in
intel_iommu_init().
- Allow NVS regions to pass RMRR check.
- Domain cleanup in error path.This series is also available at github.
https://github.com/LuBaolu/intel-iommu/commits/vtd-fix-for-v6.1-rc2
Applied, thanks Baolu.