* [PATCH v2 0/2] Change PAT to support mremap use-cases @ 2015-12-23 0:54 Toshi Kani 2015-12-23 0:54 ` [PATCH v2 1/2] x86/mm/pat: Add untrack_pfn_moved for mremap Toshi Kani 2015-12-23 0:54 ` [PATCH v2 2/2] x86/mm/pat: Change free_memtype() to support shrinking case Toshi Kani 0 siblings, 2 replies; 5+ messages in thread From: Toshi Kani @ 2015-12-23 0:54 UTC (permalink / raw) To: tglx, mingo, hpa, bp; +Cc: stsp, x86, linux-mm, linux-kernel This patch-set fixes two issues found in PAT when mremap() is used on a VM_PFNMAP range. Patch 1/2 fixes an issue for mremap() with MREMAP_FIXED, which moves a pfnmap from old vma to new vma. untrack_pfn_moved() is added to clear VM_PAT from old vma. Patch 2/2 fixes an issue for mremap() with a shrinking map size. The free_memtype() path is changed to support this mremap case in addition to the regular munmap case. Note, using mremap() with an expanded map size is not a valid case since VM_DONTEXPAND is set along with VM_PFNMAP. v2: - Add an explicit call in the mremap code which clears the PAT flag. (Thomas Gleixner) - Add comment to explain how memtype_rb_match() is used for shrinking case. (Thomas Gleixner) --- Toshi Kani (2): 1/2 x86/mm/pat: Add untrack_pfn_moved for mremap 2/2 x86/mm/pat: Change free_memtype() to support shrinking case --- arch/x86/mm/pat.c | 12 +++++++++- arch/x86/mm/pat_rbtree.c | 52 +++++++++++++++++++++++++++++++++++-------- include/asm-generic/pgtable.h | 10 ++++++++- mm/mremap.c | 4 ++++ 4 files changed, 67 insertions(+), 11 deletions(-) ^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH v2 1/2] x86/mm/pat: Add untrack_pfn_moved for mremap 2015-12-23 0:54 [PATCH v2 0/2] Change PAT to support mremap use-cases Toshi Kani @ 2015-12-23 0:54 ` Toshi Kani 2016-01-05 14:30 ` [tip:x86/mm] " tip-bot for Toshi Kani 2015-12-23 0:54 ` [PATCH v2 2/2] x86/mm/pat: Change free_memtype() to support shrinking case Toshi Kani 1 sibling, 1 reply; 5+ messages in thread From: Toshi Kani @ 2015-12-23 0:54 UTC (permalink / raw) To: tglx, mingo, hpa, bp Cc: stsp, x86, linux-mm, linux-kernel, Borislav Petkov, Toshi Kani mremap() with MREMAP_FIXED on a VM_PFNMAP range causes the following WARN_ON_ONCE() message in untrack_pfn(). WARNING: CPU: 1 PID: 3493 at arch/x86/mm/pat.c:985 untrack_pfn+0xbd/0xd0() Call Trace: [<ffffffff817729ea>] dump_stack+0x45/0x57 [<ffffffff8109e4b6>] warn_slowpath_common+0x86/0xc0 [<ffffffff8109e5ea>] warn_slowpath_null+0x1a/0x20 [<ffffffff8106a88d>] untrack_pfn+0xbd/0xd0 [<ffffffff811d2d5e>] unmap_single_vma+0x80e/0x860 [<ffffffff811d3725>] unmap_vmas+0x55/0xb0 [<ffffffff811d916c>] unmap_region+0xac/0x120 [<ffffffff811db86a>] do_munmap+0x28a/0x460 [<ffffffff811dec33>] move_vma+0x1b3/0x2e0 [<ffffffff811df113>] SyS_mremap+0x3b3/0x510 [<ffffffff817793ee>] entry_SYSCALL_64_fastpath+0x12/0x71 MREMAP_FIXED moves a pfnmap from old vma to new vma. untrack_pfn() is called with the old vma after its pfnmap page table has been removed, which causes follow_phys() to fail. The new vma has a new pfnmap to the same pfn & cache type with VM_PAT set. Therefore, we only need to clear VM_PAT from the old vma in this case. Add untrack_pfn_moved(), which clears VM_PAT from a given old vma. move_vma() is changed to call this function with the old vma when VM_PFNMAP is set. move_vma() then calls do_munmap(), and untrack_pfn() is a no-op since VM_PAT is cleared. Link: http://lkml.kernel.org/r/<1446072663.20657.150.camel@hpe.com> Reported-by: Stas Sergeev <stsp@list.ru> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Borislav Petkov <bp@suse.de> Signed-off-by: Toshi Kani <toshi.kani@hpe.com> --- arch/x86/mm/pat.c | 10 ++++++++++ include/asm-generic/pgtable.h | 10 +++++++++- mm/mremap.c | 4 ++++ 3 files changed, 23 insertions(+), 1 deletion(-) diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index 188e3e0..1aca073 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -992,6 +992,16 @@ void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn, vma->vm_flags &= ~VM_PAT; } +/* + * untrack_pfn_moved is called, while mremapping a pfnmap for a new region, + * with the old vma after its pfnmap page table has been removed. The new + * vma has a new pfnmap to the same pfn & cache type with VM_PAT set. + */ +void untrack_pfn_moved(struct vm_area_struct *vma) +{ + vma->vm_flags &= ~VM_PAT; +} + pgprot_t pgprot_writecombine(pgprot_t prot) { return __pgprot(pgprot_val(prot) | diff --git a/include/asm-generic/pgtable.h b/include/asm-generic/pgtable.h index 14b0ff32..3a6803c 100644 --- a/include/asm-generic/pgtable.h +++ b/include/asm-generic/pgtable.h @@ -569,7 +569,7 @@ static inline int track_pfn_copy(struct vm_area_struct *vma) } /* - * untrack_pfn_vma is called while unmapping a pfnmap for a region. + * untrack_pfn is called while unmapping a pfnmap for a region. * untrack can be called for a specific region indicated by pfn and size or * can be for the entire vma (in which case pfn, size are zero). */ @@ -577,6 +577,13 @@ static inline void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn, unsigned long size) { } + +/* + * untrack_pfn_moved is called while mremapping a pfnmap for a new region. + */ +static inline void untrack_pfn_moved(struct vm_area_struct *vma) +{ +} #else extern int track_pfn_remap(struct vm_area_struct *vma, pgprot_t *prot, unsigned long pfn, unsigned long addr, @@ -586,6 +593,7 @@ extern int track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot, extern int track_pfn_copy(struct vm_area_struct *vma); extern void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn, unsigned long size); +extern void untrack_pfn_moved(struct vm_area_struct *vma); #endif #ifdef __HAVE_COLOR_ZERO_PAGE diff --git a/mm/mremap.c b/mm/mremap.c index c25bc62..de824e7 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -319,6 +319,10 @@ static unsigned long move_vma(struct vm_area_struct *vma, hiwater_vm = mm->hiwater_vm; vm_stat_account(mm, vma->vm_flags, vma->vm_file, new_len>>PAGE_SHIFT); + /* Tell pfnmap has moved from this vma */ + if (unlikely(vma->vm_flags & VM_PFNMAP)) + untrack_pfn_moved(vma); + if (do_munmap(mm, old_addr, old_len) < 0) { /* OOM: unable to split vma, just get accounts right */ vm_unacct_memory(excess >> PAGE_SHIFT); ^ permalink raw reply [flat|nested] 5+ messages in thread
* [tip:x86/mm] x86/mm/pat: Add untrack_pfn_moved for mremap 2015-12-23 0:54 ` [PATCH v2 1/2] x86/mm/pat: Add untrack_pfn_moved for mremap Toshi Kani @ 2016-01-05 14:30 ` tip-bot for Toshi Kani 0 siblings, 0 replies; 5+ messages in thread From: tip-bot for Toshi Kani @ 2016-01-05 14:30 UTC (permalink / raw) To: linux-tip-commits Cc: stsp, bp, tglx, toshi.kani, hpa, mingo, linux-kernel, mingo Commit-ID: d9fe4fab11976e56b2e992980bf6ce948bdf02ac Gitweb: http://git.kernel.org/tip/d9fe4fab11976e56b2e992980bf6ce948bdf02ac Author: Toshi Kani <toshi.kani@hpe.com> AuthorDate: Tue, 22 Dec 2015 17:54:23 -0700 Committer: Thomas Gleixner <tglx@linutronix.de> CommitDate: Tue, 5 Jan 2016 11:10:05 +0100 x86/mm/pat: Add untrack_pfn_moved for mremap mremap() with MREMAP_FIXED on a VM_PFNMAP range causes the following WARN_ON_ONCE() message in untrack_pfn(). WARNING: CPU: 1 PID: 3493 at arch/x86/mm/pat.c:985 untrack_pfn+0xbd/0xd0() Call Trace: [<ffffffff817729ea>] dump_stack+0x45/0x57 [<ffffffff8109e4b6>] warn_slowpath_common+0x86/0xc0 [<ffffffff8109e5ea>] warn_slowpath_null+0x1a/0x20 [<ffffffff8106a88d>] untrack_pfn+0xbd/0xd0 [<ffffffff811d2d5e>] unmap_single_vma+0x80e/0x860 [<ffffffff811d3725>] unmap_vmas+0x55/0xb0 [<ffffffff811d916c>] unmap_region+0xac/0x120 [<ffffffff811db86a>] do_munmap+0x28a/0x460 [<ffffffff811dec33>] move_vma+0x1b3/0x2e0 [<ffffffff811df113>] SyS_mremap+0x3b3/0x510 [<ffffffff817793ee>] entry_SYSCALL_64_fastpath+0x12/0x71 MREMAP_FIXED moves a pfnmap from old vma to new vma. untrack_pfn() is called with the old vma after its pfnmap page table has been removed, which causes follow_phys() to fail. The new vma has a new pfnmap to the same pfn & cache type with VM_PAT set. Therefore, we only need to clear VM_PAT from the old vma in this case. Add untrack_pfn_moved(), which clears VM_PAT from a given old vma. move_vma() is changed to call this function with the old vma when VM_PFNMAP is set. move_vma() then calls do_munmap(), and untrack_pfn() is a no-op since VM_PAT is cleared. Reported-by: Stas Sergeev <stsp@list.ru> Signed-off-by: Toshi Kani <toshi.kani@hpe.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Borislav Petkov <bp@suse.de> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/1450832064-10093-2-git-send-email-toshi.kani@hpe.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de> --- arch/x86/mm/pat.c | 10 ++++++++++ include/asm-generic/pgtable.h | 10 +++++++++- mm/mremap.c | 4 ++++ 3 files changed, 23 insertions(+), 1 deletion(-) diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index 188e3e0..1aca073 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -992,6 +992,16 @@ void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn, vma->vm_flags &= ~VM_PAT; } +/* + * untrack_pfn_moved is called, while mremapping a pfnmap for a new region, + * with the old vma after its pfnmap page table has been removed. The new + * vma has a new pfnmap to the same pfn & cache type with VM_PAT set. + */ +void untrack_pfn_moved(struct vm_area_struct *vma) +{ + vma->vm_flags &= ~VM_PAT; +} + pgprot_t pgprot_writecombine(pgprot_t prot) { return __pgprot(pgprot_val(prot) | diff --git a/include/asm-generic/pgtable.h b/include/asm-generic/pgtable.h index 14b0ff32..3a6803c 100644 --- a/include/asm-generic/pgtable.h +++ b/include/asm-generic/pgtable.h @@ -569,7 +569,7 @@ static inline int track_pfn_copy(struct vm_area_struct *vma) } /* - * untrack_pfn_vma is called while unmapping a pfnmap for a region. + * untrack_pfn is called while unmapping a pfnmap for a region. * untrack can be called for a specific region indicated by pfn and size or * can be for the entire vma (in which case pfn, size are zero). */ @@ -577,6 +577,13 @@ static inline void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn, unsigned long size) { } + +/* + * untrack_pfn_moved is called while mremapping a pfnmap for a new region. + */ +static inline void untrack_pfn_moved(struct vm_area_struct *vma) +{ +} #else extern int track_pfn_remap(struct vm_area_struct *vma, pgprot_t *prot, unsigned long pfn, unsigned long addr, @@ -586,6 +593,7 @@ extern int track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot, extern int track_pfn_copy(struct vm_area_struct *vma); extern void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn, unsigned long size); +extern void untrack_pfn_moved(struct vm_area_struct *vma); #endif #ifdef __HAVE_COLOR_ZERO_PAGE diff --git a/mm/mremap.c b/mm/mremap.c index c25bc62..de824e7 100644 --- a/mm/mremap.c +++ b/mm/mremap.c @@ -319,6 +319,10 @@ static unsigned long move_vma(struct vm_area_struct *vma, hiwater_vm = mm->hiwater_vm; vm_stat_account(mm, vma->vm_flags, vma->vm_file, new_len>>PAGE_SHIFT); + /* Tell pfnmap has moved from this vma */ + if (unlikely(vma->vm_flags & VM_PFNMAP)) + untrack_pfn_moved(vma); + if (do_munmap(mm, old_addr, old_len) < 0) { /* OOM: unable to split vma, just get accounts right */ vm_unacct_memory(excess >> PAGE_SHIFT); ^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH v2 2/2] x86/mm/pat: Change free_memtype() to support shrinking case 2015-12-23 0:54 [PATCH v2 0/2] Change PAT to support mremap use-cases Toshi Kani 2015-12-23 0:54 ` [PATCH v2 1/2] x86/mm/pat: Add untrack_pfn_moved for mremap Toshi Kani @ 2015-12-23 0:54 ` Toshi Kani 2016-01-05 14:31 ` [tip:x86/mm] " tip-bot for Toshi Kani 1 sibling, 1 reply; 5+ messages in thread From: Toshi Kani @ 2015-12-23 0:54 UTC (permalink / raw) To: tglx, mingo, hpa, bp Cc: stsp, x86, linux-mm, linux-kernel, Borislav Petkov, Toshi Kani Using mremap() to shrink the map size of a VM_PFNMAP range causes the following error message, and leaves the pfn range allocated. x86/PAT: test:3493 freeing invalid memtype [mem 0x483200000-0x4863fffff] This is because rbt_memtype_erase(), called from free_memtype() with spin_lock held, only supports to free a whole memtype node in memtype_rbroot. Therefore, this patch changes rbt_memtype_erase() to support a request that shrinks the size of a memtype node for mremap(). memtype_rb_exact_match() is renamed to memtype_rb_match(), and is enhanced to support EXACT_MATCH and END_MATH in @match_type. Since the memtype_rbroot tree allows overlapping ranges, rbt_memtype_erase() checks with EXACT_MATCH first, i.e. free a whole node for the munmap case. If no such entry is found, it then checks with END_MATCH, i.e. shrink the size of a node from the end for the mremap case. On the mremap case, rbt_memtype_erase() proceeds in two steps, 1) remove the node, and then 2) insert the updated node. This allows proper update of augmented values, subtree_max_end, in the tree. Link: http://lkml.kernel.org/r/<1446072663.20657.150.camel@hpe.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Borislav Petkov <bp@suse.de> Signed-off-by: Toshi Kani <toshi.kani@hpe.com> --- arch/x86/mm/pat.c | 2 +- arch/x86/mm/pat_rbtree.c | 52 ++++++++++++++++++++++++++++++++++++++-------- 2 files changed, 44 insertions(+), 10 deletions(-) diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index 1aca073..031782e 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -586,7 +586,7 @@ int free_memtype(u64 start, u64 end) entry = rbt_memtype_erase(start, end); spin_unlock(&memtype_lock); - if (!entry) { + if (IS_ERR(entry)) { pr_info("x86/PAT: %s:%d freeing invalid memtype [mem %#010Lx-%#010Lx]\n", current->comm, current->pid, start, end - 1); return -EINVAL; diff --git a/arch/x86/mm/pat_rbtree.c b/arch/x86/mm/pat_rbtree.c index 6393108..2f77022 100644 --- a/arch/x86/mm/pat_rbtree.c +++ b/arch/x86/mm/pat_rbtree.c @@ -98,8 +98,13 @@ static struct memtype *memtype_rb_lowest_match(struct rb_root *root, return last_lower; /* Returns NULL if there is no overlap */ } -static struct memtype *memtype_rb_exact_match(struct rb_root *root, - u64 start, u64 end) +enum { + MEMTYPE_EXACT_MATCH = 0, + MEMTYPE_END_MATCH = 1 +}; + +static struct memtype *memtype_rb_match(struct rb_root *root, + u64 start, u64 end, int match_type) { struct memtype *match; @@ -107,7 +112,12 @@ static struct memtype *memtype_rb_exact_match(struct rb_root *root, while (match != NULL && match->start < end) { struct rb_node *node; - if (match->start == start && match->end == end) + if ((match_type == MEMTYPE_EXACT_MATCH) && + (match->start == start) && (match->end == end)) + return match; + + if ((match_type == MEMTYPE_END_MATCH) && + (match->start < start) && (match->end == end)) return match; node = rb_next(&match->rb); @@ -117,7 +127,7 @@ static struct memtype *memtype_rb_exact_match(struct rb_root *root, match = NULL; } - return NULL; /* Returns NULL if there is no exact match */ + return NULL; /* Returns NULL if there is no match */ } static int memtype_rb_check_conflict(struct rb_root *root, @@ -210,12 +220,36 @@ struct memtype *rbt_memtype_erase(u64 start, u64 end) { struct memtype *data; - data = memtype_rb_exact_match(&memtype_rbroot, start, end); - if (!data) - goto out; + /* + * Since the memtype_rbroot tree allows overlapping ranges, + * rbt_memtype_erase() checks with EXACT_MATCH first, i.e. free + * a whole node for the munmap case. If no such entry is found, + * it then checks with END_MATCH, i.e. shrink the size of a node + * from the end for the mremap case. + */ + data = memtype_rb_match(&memtype_rbroot, start, end, + MEMTYPE_EXACT_MATCH); + if (!data) { + data = memtype_rb_match(&memtype_rbroot, start, end, + MEMTYPE_END_MATCH); + if (!data) + return ERR_PTR(-EINVAL); + } + + if (data->start == start) { + /* munmap: erase this node */ + rb_erase_augmented(&data->rb, &memtype_rbroot, + &memtype_rb_augment_cb); + } else { + /* mremap: update the end value of this node */ + rb_erase_augmented(&data->rb, &memtype_rbroot, + &memtype_rb_augment_cb); + data->end = start; + data->subtree_max_end = data->end; + memtype_rb_insert(&memtype_rbroot, data); + return NULL; + } - rb_erase_augmented(&data->rb, &memtype_rbroot, &memtype_rb_augment_cb); -out: return data; } ^ permalink raw reply [flat|nested] 5+ messages in thread
* [tip:x86/mm] x86/mm/pat: Change free_memtype() to support shrinking case 2015-12-23 0:54 ` [PATCH v2 2/2] x86/mm/pat: Change free_memtype() to support shrinking case Toshi Kani @ 2016-01-05 14:31 ` tip-bot for Toshi Kani 0 siblings, 0 replies; 5+ messages in thread From: tip-bot for Toshi Kani @ 2016-01-05 14:31 UTC (permalink / raw) To: linux-tip-commits; +Cc: toshi.kani, linux-kernel, hpa, bp, mingo, mingo, tglx Commit-ID: 2039e6acaf94d83ec6b6d9f3d0bce7ea1f099918 Gitweb: http://git.kernel.org/tip/2039e6acaf94d83ec6b6d9f3d0bce7ea1f099918 Author: Toshi Kani <toshi.kani@hpe.com> AuthorDate: Tue, 22 Dec 2015 17:54:24 -0700 Committer: Thomas Gleixner <tglx@linutronix.de> CommitDate: Tue, 5 Jan 2016 11:10:23 +0100 x86/mm/pat: Change free_memtype() to support shrinking case Using mremap() to shrink the map size of a VM_PFNMAP range causes the following error message, and leaves the pfn range allocated. x86/PAT: test:3493 freeing invalid memtype [mem 0x483200000-0x4863fffff] This is because rbt_memtype_erase(), called from free_memtype() with spin_lock held, only supports to free a whole memtype node in memtype_rbroot. Therefore, this patch changes rbt_memtype_erase() to support a request that shrinks the size of a memtype node for mremap(). memtype_rb_exact_match() is renamed to memtype_rb_match(), and is enhanced to support EXACT_MATCH and END_MATCH in @match_type. Since the memtype_rbroot tree allows overlapping ranges, rbt_memtype_erase() checks with EXACT_MATCH first, i.e. free a whole node for the munmap case. If no such entry is found, it then checks with END_MATCH, i.e. shrink the size of a node from the end for the mremap case. On the mremap case, rbt_memtype_erase() proceeds in two steps, 1) remove the node, and then 2) insert the updated node. This allows proper update of augmented values, subtree_max_end, in the tree. Signed-off-by: Toshi Kani <toshi.kani@hpe.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Borislav Petkov <bp@suse.de> Cc: stsp@list.ru Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/1450832064-10093-3-git-send-email-toshi.kani@hpe.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de> --- arch/x86/mm/pat.c | 2 +- arch/x86/mm/pat_rbtree.c | 52 +++++++++++++++++++++++++++++++++++++++--------- 2 files changed, 44 insertions(+), 10 deletions(-) diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index 1aca073..031782e 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -586,7 +586,7 @@ int free_memtype(u64 start, u64 end) entry = rbt_memtype_erase(start, end); spin_unlock(&memtype_lock); - if (!entry) { + if (IS_ERR(entry)) { pr_info("x86/PAT: %s:%d freeing invalid memtype [mem %#010Lx-%#010Lx]\n", current->comm, current->pid, start, end - 1); return -EINVAL; diff --git a/arch/x86/mm/pat_rbtree.c b/arch/x86/mm/pat_rbtree.c index 6393108..2f77022 100644 --- a/arch/x86/mm/pat_rbtree.c +++ b/arch/x86/mm/pat_rbtree.c @@ -98,8 +98,13 @@ static struct memtype *memtype_rb_lowest_match(struct rb_root *root, return last_lower; /* Returns NULL if there is no overlap */ } -static struct memtype *memtype_rb_exact_match(struct rb_root *root, - u64 start, u64 end) +enum { + MEMTYPE_EXACT_MATCH = 0, + MEMTYPE_END_MATCH = 1 +}; + +static struct memtype *memtype_rb_match(struct rb_root *root, + u64 start, u64 end, int match_type) { struct memtype *match; @@ -107,7 +112,12 @@ static struct memtype *memtype_rb_exact_match(struct rb_root *root, while (match != NULL && match->start < end) { struct rb_node *node; - if (match->start == start && match->end == end) + if ((match_type == MEMTYPE_EXACT_MATCH) && + (match->start == start) && (match->end == end)) + return match; + + if ((match_type == MEMTYPE_END_MATCH) && + (match->start < start) && (match->end == end)) return match; node = rb_next(&match->rb); @@ -117,7 +127,7 @@ static struct memtype *memtype_rb_exact_match(struct rb_root *root, match = NULL; } - return NULL; /* Returns NULL if there is no exact match */ + return NULL; /* Returns NULL if there is no match */ } static int memtype_rb_check_conflict(struct rb_root *root, @@ -210,12 +220,36 @@ struct memtype *rbt_memtype_erase(u64 start, u64 end) { struct memtype *data; - data = memtype_rb_exact_match(&memtype_rbroot, start, end); - if (!data) - goto out; + /* + * Since the memtype_rbroot tree allows overlapping ranges, + * rbt_memtype_erase() checks with EXACT_MATCH first, i.e. free + * a whole node for the munmap case. If no such entry is found, + * it then checks with END_MATCH, i.e. shrink the size of a node + * from the end for the mremap case. + */ + data = memtype_rb_match(&memtype_rbroot, start, end, + MEMTYPE_EXACT_MATCH); + if (!data) { + data = memtype_rb_match(&memtype_rbroot, start, end, + MEMTYPE_END_MATCH); + if (!data) + return ERR_PTR(-EINVAL); + } + + if (data->start == start) { + /* munmap: erase this node */ + rb_erase_augmented(&data->rb, &memtype_rbroot, + &memtype_rb_augment_cb); + } else { + /* mremap: update the end value of this node */ + rb_erase_augmented(&data->rb, &memtype_rbroot, + &memtype_rb_augment_cb); + data->end = start; + data->subtree_max_end = data->end; + memtype_rb_insert(&memtype_rbroot, data); + return NULL; + } - rb_erase_augmented(&data->rb, &memtype_rbroot, &memtype_rb_augment_cb); -out: return data; } ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2016-01-05 14:31 UTC | newest] Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2015-12-23 0:54 [PATCH v2 0/2] Change PAT to support mremap use-cases Toshi Kani 2015-12-23 0:54 ` [PATCH v2 1/2] x86/mm/pat: Add untrack_pfn_moved for mremap Toshi Kani 2016-01-05 14:30 ` [tip:x86/mm] " tip-bot for Toshi Kani 2015-12-23 0:54 ` [PATCH v2 2/2] x86/mm/pat: Change free_memtype() to support shrinking case Toshi Kani 2016-01-05 14:31 ` [tip:x86/mm] " tip-bot for Toshi Kani
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox
Powered by JetHome