From ae7871be189cb41184f1e05742b4a99e2c59774d Mon Sep 17 00:00:00 2001 From: Geert Uytterhoeven Date: Fri, 16 Dec 2016 14:28:41 +0100 Subject: swiotlb: Convert swiotlb_force from int to enum Convert the flag swiotlb_force from an int to an enum, to prepare for the advent of more possible values. Suggested-by: Konrad Rzeszutek Wilk Signed-off-by: Geert Uytterhoeven Signed-off-by: Konrad Rzeszutek Wilk --- lib/swiotlb.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'lib') diff --git a/lib/swiotlb.c b/lib/swiotlb.c index cb1b54ee8527..a32dce6d5101 100644 --- a/lib/swiotlb.c +++ b/lib/swiotlb.c @@ -53,7 +53,7 @@ */ #define IO_TLB_MIN_SLABS ((1<<20) >> IO_TLB_SHIFT) -int swiotlb_force; +enum swiotlb_force swiotlb_force; /* * Used to do a quick range check in swiotlb_tbl_unmap_single and @@ -107,7 +107,7 @@ setup_io_tlb_npages(char *str) if (*str == ',') ++str; if (!strcmp(str, "force")) - swiotlb_force = 1; + swiotlb_force = SWIOTLB_FORCE; return 0; } @@ -763,7 +763,7 @@ dma_addr_t swiotlb_map_page(struct device *dev, struct page *page, * we can safely return the device addr and not worry about bounce * buffering it. */ - if (dma_capable(dev, dev_addr, size) && !swiotlb_force) + if (dma_capable(dev, dev_addr, size) && swiotlb_force != SWIOTLB_FORCE) return dev_addr; trace_swiotlb_bounced(dev, dev_addr, size, swiotlb_force); @@ -904,7 +904,7 @@ swiotlb_map_sg_attrs(struct device *hwdev, struct scatterlist *sgl, int nelems, phys_addr_t paddr = sg_phys(sg); dma_addr_t dev_addr = phys_to_dma(hwdev, paddr); - if (swiotlb_force || + if (swiotlb_force == SWIOTLB_FORCE || !dma_capable(hwdev, dev_addr, sg->length)) { phys_addr_t map = map_single(hwdev, sg_phys(sg), sg->length, dir, attrs); -- cgit v1.3-14-g43fede From fff5d99225107f5f13fe4a9805adc2a1c4b5fb00 Mon Sep 17 00:00:00 2001 From: Geert Uytterhoeven Date: Fri, 16 Dec 2016 14:28:42 +0100 Subject: swiotlb: Add swiotlb=noforce debug option On architectures like arm64, swiotlb is tied intimately to the core architecture DMA support. In addition, ZONE_DMA cannot be disabled. To aid debugging and catch devices not supporting DMA to memory outside the 32-bit address space, add a kernel command line option "swiotlb=noforce", which disables the use of bounce buffers. If specified, trying to map memory that cannot be used with DMA will fail, and a rate-limited warning will be printed. Note that io_tlb_nslabs is set to 1, which is the minimal supported value. Signed-off-by: Geert Uytterhoeven Signed-off-by: Konrad Rzeszutek Wilk --- Documentation/admin-guide/kernel-parameters.txt | 3 ++- include/linux/swiotlb.h | 1 + include/trace/events/swiotlb.h | 3 ++- lib/swiotlb.c | 18 ++++++++++++++++-- 4 files changed, 21 insertions(+), 4 deletions(-) (limited to 'lib') diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt index be2d6d0a03a4..7d5f4f4cecfa 100644 --- a/Documentation/admin-guide/kernel-parameters.txt +++ b/Documentation/admin-guide/kernel-parameters.txt @@ -3807,10 +3807,11 @@ it if 0 is given (See Documentation/cgroup-v1/memory.txt) swiotlb= [ARM,IA-64,PPC,MIPS,X86] - Format: { | force } + Format: { | force | noforce } -- Number of I/O TLB slabs force -- force using of bounce buffers even if they wouldn't be automatically used by the kernel + noforce -- Never use bounce buffers (for debugging) switches= [HW,M68k] diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h index 71d104e4c849..d9c84a9cde3d 100644 --- a/include/linux/swiotlb.h +++ b/include/linux/swiotlb.h @@ -12,6 +12,7 @@ struct scatterlist; enum swiotlb_force { SWIOTLB_NORMAL, /* Default - depending on HW DMA mask etc. */ SWIOTLB_FORCE, /* swiotlb=force */ + SWIOTLB_NO_FORCE, /* swiotlb=noforce */ }; extern enum swiotlb_force swiotlb_force; diff --git a/include/trace/events/swiotlb.h b/include/trace/events/swiotlb.h index 5e2e30a7efce..288c0c54a2b4 100644 --- a/include/trace/events/swiotlb.h +++ b/include/trace/events/swiotlb.h @@ -39,7 +39,8 @@ TRACE_EVENT(swiotlb_bounced, __entry->size, __print_symbolic(__entry->swiotlb_force, { SWIOTLB_NORMAL, "NORMAL" }, - { SWIOTLB_FORCE, "FORCE" })) + { SWIOTLB_FORCE, "FORCE" }, + { SWIOTLB_NO_FORCE, "NO_FORCE" })) ); #endif /* _TRACE_SWIOTLB_H */ diff --git a/lib/swiotlb.c b/lib/swiotlb.c index a32dce6d5101..9def738af4f4 100644 --- a/lib/swiotlb.c +++ b/lib/swiotlb.c @@ -106,8 +106,12 @@ setup_io_tlb_npages(char *str) } if (*str == ',') ++str; - if (!strcmp(str, "force")) + if (!strcmp(str, "force")) { swiotlb_force = SWIOTLB_FORCE; + } else if (!strcmp(str, "noforce")) { + swiotlb_force = SWIOTLB_NO_FORCE; + io_tlb_nslabs = 1; + } return 0; } @@ -543,8 +547,15 @@ static phys_addr_t map_single(struct device *hwdev, phys_addr_t phys, size_t size, enum dma_data_direction dir, unsigned long attrs) { - dma_addr_t start_dma_addr = phys_to_dma(hwdev, io_tlb_start); + dma_addr_t start_dma_addr; + + if (swiotlb_force == SWIOTLB_NO_FORCE) { + dev_warn_ratelimited(hwdev, "Cannot do DMA to address %pa\n", + &phys); + return SWIOTLB_MAP_ERROR; + } + start_dma_addr = phys_to_dma(hwdev, io_tlb_start); return swiotlb_tbl_map_single(hwdev, start_dma_addr, phys, size, dir, attrs); } @@ -721,6 +732,9 @@ static void swiotlb_full(struct device *dev, size_t size, enum dma_data_direction dir, int do_panic) { + if (swiotlb_force == SWIOTLB_NO_FORCE) + return; + /* * Ran out of IOMMU space for this operation. This is very bad. * Unfortunately the drivers cannot handle this operation properly. -- cgit v1.3-14-g43fede From 7453c549f5f6485c0d79cad7844870dcc7d1b34d Mon Sep 17 00:00:00 2001 From: Konrad Rzeszutek Wilk Date: Tue, 20 Dec 2016 10:02:02 -0500 Subject: swiotlb: Export swiotlb_max_segment to users So they can figure out what is the optimal number of pages that can be contingously stitched together without fear of bounce buffer. We also expose an mechanism for sub-users of SWIOTLB API, such as Xen-SWIOTLB to set the max segment value. And lastly if swiotlb=force is set (which mandates we bounce buffer everything) we set max_segment so at least we can bounce buffer one 4K page instead of a giant 512KB one for which we may not have space. Signed-off-by: Konrad Rzeszutek Wilk Reported-and-Tested-by: Juergen Gross --- drivers/gpu/drm/i915/i915_gem.c | 11 +---------- drivers/xen/swiotlb-xen.c | 4 ++++ include/linux/swiotlb.h | 3 +++ lib/swiotlb.c | 26 ++++++++++++++++++++++++++ 4 files changed, 34 insertions(+), 10 deletions(-) (limited to 'lib') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index d0dcaf35b429..115fa399608c 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2290,15 +2290,6 @@ unlock: mutex_unlock(&obj->mm.lock); } -static unsigned int swiotlb_max_size(void) -{ -#if IS_ENABLED(CONFIG_SWIOTLB) - return rounddown(swiotlb_nr_tbl() << IO_TLB_SHIFT, PAGE_SIZE); -#else - return 0; -#endif -} - static void i915_sg_trim(struct sg_table *orig_st) { struct sg_table new_st; @@ -2345,7 +2336,7 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj) GEM_BUG_ON(obj->base.read_domains & I915_GEM_GPU_DOMAINS); GEM_BUG_ON(obj->base.write_domain & I915_GEM_GPU_DOMAINS); - max_segment = swiotlb_max_size(); + max_segment = swiotlb_max_segment(); if (!max_segment) max_segment = rounddown(UINT_MAX, PAGE_SIZE); diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c index aba12009422e..f905d6eeb048 100644 --- a/drivers/xen/swiotlb-xen.c +++ b/drivers/xen/swiotlb-xen.c @@ -275,6 +275,10 @@ retry: rc = 0; } else rc = swiotlb_late_init_with_tbl(xen_io_tlb_start, xen_io_tlb_nslabs); + + if (!rc) + swiotlb_set_max_segment(PAGE_SIZE); + return rc; error: if (repeat--) { diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h index d9c84a9cde3d..4ee479f2f355 100644 --- a/include/linux/swiotlb.h +++ b/include/linux/swiotlb.h @@ -114,11 +114,14 @@ swiotlb_dma_supported(struct device *hwdev, u64 mask); #ifdef CONFIG_SWIOTLB extern void __init swiotlb_free(void); +unsigned int swiotlb_max_segment(void); #else static inline void swiotlb_free(void) { } +static inline unsigned int swiotlb_max_segment(void) { return 0; } #endif extern void swiotlb_print_info(void); extern int is_swiotlb_buffer(phys_addr_t paddr); +extern void swiotlb_set_max_segment(unsigned int); #endif /* __LINUX_SWIOTLB_H */ diff --git a/lib/swiotlb.c b/lib/swiotlb.c index 9def738af4f4..975b8fc4f1e1 100644 --- a/lib/swiotlb.c +++ b/lib/swiotlb.c @@ -82,6 +82,12 @@ static phys_addr_t io_tlb_overflow_buffer; static unsigned int *io_tlb_list; static unsigned int io_tlb_index; +/* + * Max segment that we can provide which (if pages are contingous) will + * not be bounced (unless SWIOTLB_FORCE is set). + */ +unsigned int max_segment; + /* * We need to save away the original address corresponding to a mapped entry * for the sync operations. @@ -124,6 +130,20 @@ unsigned long swiotlb_nr_tbl(void) } EXPORT_SYMBOL_GPL(swiotlb_nr_tbl); +unsigned int swiotlb_max_segment(void) +{ + return max_segment; +} +EXPORT_SYMBOL_GPL(swiotlb_max_segment); + +void swiotlb_set_max_segment(unsigned int val) +{ + if (swiotlb_force == SWIOTLB_FORCE) + max_segment = 1; + else + max_segment = rounddown(val, PAGE_SIZE); +} + /* default to 64MB */ #define IO_TLB_DEFAULT_SIZE (64UL<<20) unsigned long swiotlb_size_or_default(void) @@ -205,6 +225,7 @@ int __init swiotlb_init_with_tbl(char *tlb, unsigned long nslabs, int verbose) if (verbose) swiotlb_print_info(); + swiotlb_set_max_segment(io_tlb_nslabs << IO_TLB_SHIFT); return 0; } @@ -283,6 +304,7 @@ swiotlb_late_init_with_default_size(size_t default_size) rc = swiotlb_late_init_with_tbl(vstart, io_tlb_nslabs); if (rc) free_pages((unsigned long)vstart, order); + return rc; } @@ -337,6 +359,8 @@ swiotlb_late_init_with_tbl(char *tlb, unsigned long nslabs) late_alloc = 1; + swiotlb_set_max_segment(io_tlb_nslabs << IO_TLB_SHIFT); + return 0; cleanup4: @@ -351,6 +375,7 @@ cleanup2: io_tlb_end = 0; io_tlb_start = 0; io_tlb_nslabs = 0; + max_segment = 0; return -ENOMEM; } @@ -379,6 +404,7 @@ void __init swiotlb_free(void) PAGE_ALIGN(io_tlb_nslabs << IO_TLB_SHIFT)); } io_tlb_nslabs = 0; + max_segment = 0; } int is_swiotlb_buffer(phys_addr_t paddr) -- cgit v1.3-14-g43fede From ea07b862ac8ef9b8c8358517d2e39f847dda6659 Mon Sep 17 00:00:00 2001 From: Johannes Weiner Date: Fri, 6 Jan 2017 19:21:43 -0500 Subject: mm: workingset: fix use-after-free in shadow node shrinker Several people report seeing warnings about inconsistent radix tree nodes followed by crashes in the workingset code, which all looked like use-after-free access from the shadow node shrinker. Dave Jones managed to reproduce the issue with a debug patch applied, which confirmed that the radix tree shrinking indeed frees shadow nodes while they are still linked to the shadow LRU: WARNING: CPU: 2 PID: 53 at lib/radix-tree.c:643 delete_node+0x1e4/0x200 CPU: 2 PID: 53 Comm: kswapd0 Not tainted 4.10.0-rc2-think+ #3 Call Trace: delete_node+0x1e4/0x200 __radix_tree_delete_node+0xd/0x10 shadow_lru_isolate+0xe6/0x220 __list_lru_walk_one.isra.4+0x9b/0x190 list_lru_walk_one+0x23/0x30 scan_shadow_nodes+0x2e/0x40 shrink_slab.part.44+0x23d/0x5d0 shrink_node+0x22c/0x330 kswapd+0x392/0x8f0 This is the WARN_ON_ONCE(!list_empty(&node->private_list)) placed in the inlined radix_tree_shrink(). The problem is with 14b468791fa9 ("mm: workingset: move shadow entry tracking to radix tree exceptional tracking"), which passes an update callback into the radix tree to link and unlink shadow leaf nodes when tree entries change, but forgot to pass the callback when reclaiming a shadow node. While the reclaimed shadow node itself is unlinked by the shrinker, its deletion from the tree can cause the left-most leaf node in the tree to be shrunk. If that happens to be a shadow node as well, we don't unlink it from the LRU as we should. Consider this tree, where the s are shadow entries: root->rnode | [0 n] | | [s ] [sssss] Now the shadow node shrinker reclaims the rightmost leaf node through the shadow node LRU: root->rnode | [0 ] | [s ] Because the parent of the deleted node is the first level below the root and has only one child in the left-most slot, the intermediate level is shrunk and the node containing the single shadow is put in its place: root->rnode | [s ] The shrinker again sees a single left-most slot in a first level node and thus decides to store the shadow in root->rnode directly and free the node - which is a leaf node on the shadow node LRU. root->rnode | s Without the update callback, the freed node remains on the shadow LRU, where it causes later shrinker runs to crash. Pass the node updater callback into __radix_tree_delete_node() in case the deletion causes the left-most branch in the tree to collapse too. Also add warnings when linked nodes are freed right away, rather than wait for the use-after-free when the list is scanned much later. Fixes: 14b468791fa9 ("mm: workingset: move shadow entry tracking to radix tree exceptional tracking") Reported-by: Dave Chinner Reported-by: Hugh Dickins Reported-by: Andrea Arcangeli Reported-and-tested-by: Dave Jones Signed-off-by: Johannes Weiner Cc: Christoph Hellwig Cc: Chris Leech Cc: Lee Duncan Cc: Jan Kara Cc: Kirill A. Shutemov Cc: Matthew Wilcox Cc: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/radix-tree.h | 4 +++- lib/radix-tree.c | 11 +++++++++-- mm/workingset.c | 3 ++- 3 files changed, 14 insertions(+), 4 deletions(-) (limited to 'lib') diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h index 5dea8f6440e4..52bda854593b 100644 --- a/include/linux/radix-tree.h +++ b/include/linux/radix-tree.h @@ -306,7 +306,9 @@ void radix_tree_iter_replace(struct radix_tree_root *, void radix_tree_replace_slot(struct radix_tree_root *root, void **slot, void *item); void __radix_tree_delete_node(struct radix_tree_root *root, - struct radix_tree_node *node); + struct radix_tree_node *node, + radix_tree_update_node_t update_node, + void *private); void *radix_tree_delete_item(struct radix_tree_root *, unsigned long, void *); void *radix_tree_delete(struct radix_tree_root *, unsigned long); void radix_tree_clear_tags(struct radix_tree_root *root, diff --git a/lib/radix-tree.c b/lib/radix-tree.c index 6f382e07de77..0b92d605fb69 100644 --- a/lib/radix-tree.c +++ b/lib/radix-tree.c @@ -640,6 +640,7 @@ static inline void radix_tree_shrink(struct radix_tree_root *root, update_node(node, private); } + WARN_ON_ONCE(!list_empty(&node->private_list)); radix_tree_node_free(node); } } @@ -666,6 +667,7 @@ static void delete_node(struct radix_tree_root *root, root->rnode = NULL; } + WARN_ON_ONCE(!list_empty(&node->private_list)); radix_tree_node_free(node); node = parent; @@ -767,6 +769,7 @@ static void radix_tree_free_nodes(struct radix_tree_node *node) struct radix_tree_node *old = child; offset = child->offset + 1; child = child->parent; + WARN_ON_ONCE(!list_empty(&node->private_list)); radix_tree_node_free(old); if (old == entry_to_node(node)) return; @@ -1824,15 +1827,19 @@ EXPORT_SYMBOL(radix_tree_gang_lookup_tag_slot); * __radix_tree_delete_node - try to free node after clearing a slot * @root: radix tree root * @node: node containing @index + * @update_node: callback for changing leaf nodes + * @private: private data to pass to @update_node * * After clearing the slot at @index in @node from radix tree * rooted at @root, call this function to attempt freeing the * node and shrinking the tree. */ void __radix_tree_delete_node(struct radix_tree_root *root, - struct radix_tree_node *node) + struct radix_tree_node *node, + radix_tree_update_node_t update_node, + void *private) { - delete_node(root, node, NULL, NULL); + delete_node(root, node, update_node, private); } /** diff --git a/mm/workingset.c b/mm/workingset.c index 241fa5d6b3b2..abb58ffa3c64 100644 --- a/mm/workingset.c +++ b/mm/workingset.c @@ -473,7 +473,8 @@ static enum lru_status shadow_lru_isolate(struct list_head *item, if (WARN_ON_ONCE(node->exceptional)) goto out_invalid; inc_node_state(page_pgdat(virt_to_page(node)), WORKINGSET_NODERECLAIM); - __radix_tree_delete_node(&mapping->page_tree, node); + __radix_tree_delete_node(&mapping->page_tree, node, + workingset_update_node, mapping); out_invalid: spin_unlock(&mapping->tree_lock); -- cgit v1.3-14-g43fede