The page size used to map the namespace is arch dependent. For example architectures like ppc64 use 16MB page size for direct-mapping. If the namespace size is not aligned to the mapping page size, we can observe kernel crash during namespace init and destroy. This is due to kernel doing partial map/unmap of the resource range BUG: Unable to handle kernel data access at 0xc001000406000000 Faulting instruction address: 0xc000000000090790 NIP [c000000000090790] arch_add_memory+0xc0/0x130 LR [c000000000090744] arch_add_memory+0x74/0x130 Call Trace: arch_add_memory+0x74/0x130 (unreliable) memremap_pages+0x74c/0xa30 devm_memremap_pages+0x3c/0xa0 pmem_attach_disk+0x188/0x770 nvdimm_bus_probe+0xd8/0x470 really_probe+0x148/0x570 driver_probe_device+0x19c/0x1d0 device_driver_attach+0xcc/0x100 bind_store+0x134/0x1c0 drv_attr_store+0x44/0x60 sysfs_kf_write+0x74/0xc0 kernfs_fop_write+0x1b4/0x290 __vfs_write+0x3c/0x70 vfs_write+0xd0/0x260 ksys_write+0xdc/0x130 system_call+0x5c/0x68 Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> --- arch/arm64/mm/flush.c | 6 ++++++ arch/powerpc/lib/pmem.c | 10 ++++++++++ arch/x86/mm/pageattr.c | 7 +++++++ include/linux/libnvdimm.h | 1 + 4 files changed, 24 insertions(+) diff --git a/arch/arm64/mm/flush.c b/arch/arm64/mm/flush.c index ac485163a4a7..36b6f6353242 100644 --- a/arch/arm64/mm/flush.c +++ b/arch/arm64/mm/flush.c @@ -91,4 +91,10 @@ void arch_invalidate_pmem(void *addr, size_t size) __inval_dcache_area(addr, size); } EXPORT_SYMBOL_GPL(arch_invalidate_pmem); + +unsigned long arch_namespace_align_size(void) +{ + return PAGE_SIZE; +} +EXPORT_SYMBOL_GPL(arch_namespace_align_size); #endif diff --git a/arch/powerpc/lib/pmem.c b/arch/powerpc/lib/pmem.c index 0666a8d29596..8f09d94b1469 100644 --- a/arch/powerpc/lib/pmem.c +++ b/arch/powerpc/lib/pmem.c @@ -26,6 +26,16 @@ void arch_invalidate_pmem(void *addr, size_t size) } EXPORT_SYMBOL_GPL(arch_invalidate_pmem); +unsigned long arch_namespace_align_size(void) +{ + + if (radix_enabled()) + return PAGE_SIZE; + return (1UL << mmu_psize_defs[mmu_linear_psize].shift); + +} +EXPORT_SYMBOL_GPL(arch_namespace_align_size); + /* * CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE symbols */ diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c index 0d09cc5aad61..51b8d11e7b94 100644 --- a/arch/x86/mm/pageattr.c +++ b/arch/x86/mm/pageattr.c @@ -310,6 +310,13 @@ void arch_invalidate_pmem(void *addr, size_t size) } EXPORT_SYMBOL_GPL(arch_invalidate_pmem); +unsigned long arch_namespace_align_size(void) +{ + return PAGE_SIZE; +} +EXPORT_SYMBOL_GPL(arch_namespace_align_size); + + static void __cpa_flush_all(void *arg) { unsigned long cache = (unsigned long)arg; diff --git a/include/linux/libnvdimm.h b/include/linux/libnvdimm.h index b6eddf912568..72d583fa9954 100644 --- a/include/linux/libnvdimm.h +++ b/include/linux/libnvdimm.h @@ -291,4 +291,5 @@ static inline void arch_invalidate_pmem(void *addr, size_t size) } #endif +unsigned long arch_namespace_align_size(void); #endif /* __LIBNVDIMM_H__ */ -- 2.23.0 _______________________________________________ Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org To unsubscribe send an email to linux-nvdimm-leave@lists.01.org
Make sure namespace start addr and size are properly aligned as per architecture restrictions. If the namespace is not aligned we mark the namespace 'disabled' Architectures like ppc64 use different page size than PAGE_SIZE to map direct-map address range. The kernel needs to make sure the namespace size is aligned correctly for the direct-map page size. kernel log will contain information as below. [ 5.810939] nd_pmem namespace0.1: invalid size/SPA [ 5.810969] nd_pmem: probe of namespace0.1 failed with error -95 and the namespace will be marked 'disabled' { "dev":"namespace0.1", "mode":"fsdax", "map":"mem", "size":1071644672, "uuid":"25577a00-c012-421d-89ca-3ee189e08848", "sector_size":512, "state":"disabled" }, Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> --- drivers/nvdimm/namespace_devs.c | 39 +++++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+) diff --git a/drivers/nvdimm/namespace_devs.c b/drivers/nvdimm/namespace_devs.c index cca0a3ba1d2c..d77d9c9e449d 100644 --- a/drivers/nvdimm/namespace_devs.c +++ b/drivers/nvdimm/namespace_devs.c @@ -1139,6 +1139,40 @@ resource_size_t nvdimm_namespace_capacity(struct nd_namespace_common *ndns) } EXPORT_SYMBOL(nvdimm_namespace_capacity); +static bool nvdimm_valid_namespace(struct nd_namespace_common *ndns, resource_size_t size) +{ + struct device *dev = &ndns->dev; + struct nd_region *nd_region = to_nd_region(dev->parent); + unsigned long align_size = arch_namespace_align_size(); + struct resource *res; + u32 remainder; + + if (is_namespace_blk(dev)) + return true; + + div_u64_rem(size, align_size * nd_region->ndr_mappings, &remainder); + if (remainder) + return false; + + if (is_namespace_pmem(dev)) { + struct nd_namespace_pmem *nspm = to_nd_namespace_pmem(dev); + + res = &nspm->nsio.res; + } else if (is_namespace_io(dev)) { + struct nd_namespace_io *nsio = to_nd_namespace_io(dev); + + res = &nsio->res; + } else + /* cannot reach */ + return false; + + div_u64_rem(res->start, align_size * nd_region->ndr_mappings, &remainder); + if (remainder) + return false; + + return true; +} + bool nvdimm_namespace_locked(struct nd_namespace_common *ndns) { int i; @@ -1735,6 +1769,11 @@ struct nd_namespace_common *nvdimm_namespace_common_probe(struct device *dev) return ERR_PTR(-ENODEV); } + if (!nvdimm_valid_namespace(ndns, size)) { + dev_err(&ndns->dev, "invalid size/SPA"); + return ERR_PTR(-EOPNOTSUPP); + } + if (is_namespace_pmem(&ndns->dev)) { struct nd_namespace_pmem *nspm; -- 2.23.0 _______________________________________________ Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org To unsubscribe send an email to linux-nvdimm-leave@lists.01.org
Architectures like ppc64 use different page size than PAGE_SIZE to map direct-map address range. The kernel needs to make sure the namespace size is aligned correctly for the direct-map page size. kernel log will contain the below details [ 939.620064] nd namespace0.3: 1071644672 is not 16384K aligned Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> --- drivers/nvdimm/namespace_devs.c | 21 ++++++++++++++++----- 1 file changed, 16 insertions(+), 5 deletions(-) diff --git a/drivers/nvdimm/namespace_devs.c b/drivers/nvdimm/namespace_devs.c index d77d9c9e449d..0e4c04765cb8 100644 --- a/drivers/nvdimm/namespace_devs.c +++ b/drivers/nvdimm/namespace_devs.c @@ -917,6 +917,17 @@ static int grow_dpa_allocation(struct nd_region *nd_region, return 0; } +static unsigned long nvdimm_validate_namespace_size(struct nd_region *nd_region, + unsigned long size, unsigned long align_size) +{ + u32 remainder; + + div_u64_rem(size, align_size * nd_region->ndr_mappings, &remainder); + if (remainder) + return align_size * nd_region->ndr_mappings; + return 0; +} + static void nd_namespace_pmem_set_resource(struct nd_region *nd_region, struct nd_namespace_pmem *nspm, resource_size_t size) { @@ -975,7 +986,8 @@ static ssize_t __size_store(struct device *dev, unsigned long long val) struct nd_mapping *nd_mapping; struct nvdimm_drvdata *ndd; struct nd_label_id label_id; - u32 flags = 0, remainder; + unsigned long map_size; + u32 flags = 0; int rc, i, id = -1; u8 *uuid = NULL; @@ -1006,10 +1018,9 @@ static ssize_t __size_store(struct device *dev, unsigned long long val) return -ENXIO; } - div_u64_rem(val, PAGE_SIZE * nd_region->ndr_mappings, &remainder); - if (remainder) { - dev_dbg(dev, "%llu is not %ldK aligned\n", val, - (PAGE_SIZE * nd_region->ndr_mappings) / SZ_1K); + map_size = nvdimm_validate_namespace_size(nd_region, val, arch_namespace_align_size()); + if (map_size) { + dev_err(dev, "%llu is not %ldK aligned\n", val, map_size / SZ_1K); return -EINVAL; } -- 2.23.0 _______________________________________________ Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org To unsubscribe send an email to linux-nvdimm-leave@lists.01.org
This should enable us to catch if we are initializing the namespace with a wrong size. Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> --- drivers/nvdimm/namespace_devs.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/drivers/nvdimm/namespace_devs.c b/drivers/nvdimm/namespace_devs.c index 0e4c04765cb8..06b55c41660d 100644 --- a/drivers/nvdimm/namespace_devs.c +++ b/drivers/nvdimm/namespace_devs.c @@ -967,6 +967,18 @@ static void nd_namespace_pmem_set_resource(struct nd_region *nd_region, out: res->start = nd_region->ndr_start + offset; res->end = res->start + size - 1; +#ifdef CONFIG_DEBUG_VM + if (size) { + unsigned long map_size; + + map_size = nvdimm_validate_namespace_size(nd_region, size, arch_namespace_align_size()); + WARN_ON(map_size); + + map_size = nvdimm_validate_namespace_size(nd_region, res->start, arch_namespace_align_size()); + WARN_ON(map_size); + + } +#endif } static bool uuid_not_set(const u8 *uuid, struct device *dev, const char *where) -- 2.23.0 _______________________________________________ Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org To unsubscribe send an email to linux-nvdimm-leave@lists.01.org
When creating new namespace make sure DPA address are properly aligned based on arch restrictions. Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> --- drivers/nvdimm/namespace_devs.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/nvdimm/namespace_devs.c b/drivers/nvdimm/namespace_devs.c index 06b55c41660d..6a83db9f8734 100644 --- a/drivers/nvdimm/namespace_devs.c +++ b/drivers/nvdimm/namespace_devs.c @@ -589,6 +589,8 @@ static void space_valid(struct nd_region *nd_region, struct nvdimm_drvdata *ndd, return; } + valid->start = ALIGN(valid->start, arch_namespace_align_size()); + /* allocation needs to be contiguous, so this is all or nothing */ if (resource_size(valid) < n) goto invalid; -- 2.23.0 _______________________________________________ Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org To unsubscribe send an email to linux-nvdimm-leave@lists.01.org
Expose supported size align as a namespace RO attribute. Usespace can use this to validate the size argument specified when creating a new namespace. Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> --- drivers/nvdimm/namespace_devs.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/drivers/nvdimm/namespace_devs.c b/drivers/nvdimm/namespace_devs.c index 6a83db9f8734..22c714469f87 100644 --- a/drivers/nvdimm/namespace_devs.c +++ b/drivers/nvdimm/namespace_devs.c @@ -1684,6 +1684,14 @@ static ssize_t force_raw_show(struct device *dev, } static DEVICE_ATTR_RW(force_raw); +static ssize_t supported_size_align_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "%ld\n", arch_namespace_align_size()); +} +static DEVICE_ATTR_RO(supported_size_align); + + static struct attribute *nd_namespace_attributes[] = { &dev_attr_nstype.attr, &dev_attr_size.attr, @@ -1696,6 +1704,7 @@ static struct attribute *nd_namespace_attributes[] = { &dev_attr_sector_size.attr, &dev_attr_dpa_extents.attr, &dev_attr_holder_class.attr, + &dev_attr_supported_size_align.attr, NULL, }; -- 2.23.0 _______________________________________________ Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org To unsubscribe send an email to linux-nvdimm-leave@lists.01.org