coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping
Subsystems (2)
RFC PATCH 1/1 coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping
From: Andrea Arcangeli <[email protected]>
Upstream 04f5866e41fb70690e28397487d8bd8eea7d712a commit.
The core dumping code has always run without holding the mmap_sem for writing, despite that is the only way to ensure that the entire vma layout will not change from under it. Only using some signal serialization on the processes belonging to the mm is not nearly enough. This was pointed out earlier. For example in Hugh's post from Jul 2017:
https://lkml.kernel.org/r/[email protected]
"Not strictly relevant here, but a related note: I was very surprised
to discover, only quite recently, how handle_mm_fault() may be called without down_read(mmap_sem) - when core dumping. That seems a misguided optimization to me, which would also be nice to correct"
In particular because the growsdown and growsup can move the vm_start/vm_end the various loops the core dump does around the vma will not be consistent if page faults can happen concurrently.
Pretty much all users calling mmget_not_zero()/get_task_mm() and then taking the mmap_sem had the potential to introduce unexpected side effects in the core dumping code.
Adding mmap_sem for writing around the ->core_dump invocation is a viable long term fix, but it requires removing all copy user and page faults and to replace them with get_dump_page() for all binary formats which is not suitable as a short term fix.
For the time being this solution manually covers the places that can confuse the core dump either by altering the vma layout or the vma flags while it runs. Once ->core_dump runs under mmap_sem for writing the function mmget_still_valid() can be dropped.
Allowing mmap_sem protected sections to run in parallel with the coredump provides some minor parallelism advantage to the swapoff code (which seems to be safe enough by never mangling any vma field and can keep doing swapins in parallel to the core dumping) and to some other corner case.
In order to facilitate the backporting I added "Fixes: 86039bd3b4e6" however the side effect of this same race condition in /proc/pid/mem should be reproducible since before 2.6.12-rc2 so I couldn't add any other "Fixes:" because there's no hash beyond the git genesis commit.
Because find_extend_vma() is the only location outside of the process context that could modify the "mm" structures under mmap_sem for reading, by adding the mmget_still_valid() check to it, all other cases that take the mmap_sem for reading don't need the new check after mmget_not_zero()/get_task_mm(). The expand_stack() in page fault context also doesn't need the new check, because all tasks under core dumping are frozen.
[[email protected]: stable 4.4 backport
- drop infiniband part because of missing 5f9794dc94f59
- drop userfaultfd_event_wait_completion hunk because of missing 9cd75c3cd4c3d]
- handle binder_update_page_range because of missing 720c241924046
- add check to collapse_huge_page
]
---
Hi,
this is based on the backport I have done for out 4.4 based distribution kernel. Please double check that I haven't missed anything before applying to the stable tree. I have also CCed Joel for the binder part which is not in the current upstream anymore but I believe it needs the check as well.
Review feedback welcome.
drivers/android/binder.c | 6 ++++++ fs/proc/task_mmu.c | 18 ++++++++++++++++++ fs/userfaultfd.c | 10 ++++++++-- include/linux/mm.h | 21 +++++++++++++++++++++ mm/huge_memory.c | 2 +- mm/mmap.c | 7 ++++++- 6 files changed, 60 insertions(+), 4 deletions(-)
drivers/android/binder.c +6 -0
@@ -570,6 +570,12 @@ static int binder_update_page_range(struct binder_proc *proc, int allocate,if (mm) {down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) { +if (allocate ==0 ) + goto free_range; + goto err_no_vma; + } + vma = proc->vma;if (vma && mm != proc->vma_vm_mm) { pr_err("%d: vma mm and task mm mismatch\n",
fs/proc/task_mmu.c +18 -0
@@ -947,6 +947,24 @@ static ssize_t clear_refs_write(struct file *file, const char __user *buf,continue ;up_read (&mm->mmap_sem);down_write (&mm->mmap_sem); +/* +* Avoid to modify vma->vm_flags +* without locked ops while the +* coredump reads the vm_flags. +*/ +if (!mmget_still_valid (mm)) { +/* +* Silently return "count" +* like if get_task_mm() +* failed. FIXME: should this +* function have returned +* -ESRCH if get_task_mm() +* failed like if +* get_proc_task() fails? +*/ +up_write (&mm->mmap_sem); + goto out_mm; + }for (vma = mm->mmap; vma; vma = vma->vm_next) { vma->vm_flags &= ~VM_SOFTDIRTY ;vma_set_page_prot (vma);
fs/userfaultfd.c +8 -2
@@ -443,6 +443,8 @@ static int userfaultfd_release(struct inode *inode, struct file *file)* taking the mmap_sem for writing. */ down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) + goto skip_mm; prev =NULL ;for (vma = mm->mmap; vma; vma = vma->vm_next) {cond_resched ();
@@ -465,6 +467,7 @@ static int userfaultfd_release(struct inode *inode, struct file *file) vma->vm_flags = new_flags; vma->vm_userfaultfd_ctx =NULL_VM_UFFD_CTX ; } +skip_mm:up_write (&mm->mmap_sem);/*
@@ -761,9 +764,10 @@ static int userfaultfd_register(struct userfaultfd_ctx *ctx, end = start + uffdio_register.range.len;down_write (&mm->mmap_sem); - vma =find_vma_prev (mm, start, &prev); - ret = -ENOMEM ; +if (!mmget_still_valid (mm)) + goto out_unlock; + vma =find_vma_prev (mm, start, &prev);if (!vma) goto out_unlock;
@@ -903,6 +907,8 @@ static int userfaultfd_unregister(struct userfaultfd_ctx *ctx, end = start + uffdio_unregister.len;down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) + goto out_unlock; vma =find_vma_prev (mm, start, &prev); ret = -ENOMEM ;
include/linux/mm.h +21 -0
@@ -1098,6 +1098,27 @@ void zap_page_range(struct vm_area_struct *vma, unsigned long address,void unmap_vmas (struct mmu_gather *tlb,struct vm_area_struct *start_vma,unsigned long start,unsigned long end); +/* +* This has to be called after a get_task_mm()/mmget_not_zero() +* followed by taking the mmap_sem for writing before modifying the +* vmas or anything the coredump pretends not to change from under it. +* +* NOTE: find_extend_vma() called from GUP context is the only place +* that can modify the "mm" (notably the vm_start/end) under mmap_sem +* for reading and outside the context of the process, so it is also +* the only case that holds the mmap_sem for reading that must call +* this function. Generally if the mmap_sem is hold for reading +* there's no need of this check after get_task_mm()/mmget_not_zero(). +* +* This function can be obsoleted and the check can be removed, after +* the coredump code will hold the mmap_sem for writing before +* invoking the ->core_dump methods. +*/ +static inline bool mmget_still_valid (struct mm_struct *mm) +{ +return likely (!mm->core_state); +} +/** * mm_walk - callbacks for walk_page_range * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
mm/huge_memory.c +1 -1
@@ -2587,7 +2587,7 @@ static void collapse_huge_page(struct mm_struct *mm,* handled by the anon_vma lock + PG_lock. */ down_write (&mm->mmap_sem); -if (unlikely (khugepaged_test_exit (mm))) +if (unlikely (khugepaged_test_exit (mm) ||mmget_still_valid (mm))) goto out; vma =find_vma (mm, address);
mm/mmap.c +6 -1
@@ -42,6 +42,7 @@#include <linux/memory.h> #include <linux/printk.h> #include <linux/userfaultfd_k.h> +#include <linux/mm.h> #include <asm/uaccess.h> #include <asm/cacheflush.h>
@@ -2398,7 +2399,8 @@ find_extend_vma(struct mm_struct *mm, unsigned long addr) vma =find_vma_prev (mm, addr, &prev);if (vma && (vma->vm_start <= addr))return vma; -if (!prev ||expand_stack (prev, addr)) +/* don't alter vm_end if the coredump is running */ +if (!prev || !mmget_still_valid (mm) ||expand_stack (prev, addr))return NULL ;if (prev->vm_flags &VM_LOCKED )populate_vma_page_range (prev, addr, prev->vm_end,NULL );
@@ -2424,6 +2426,9 @@ find_extend_vma(struct mm_struct *mm, unsigned long addr)return vma;if (!(vma->vm_flags &VM_GROWSDOWN ))return NULL ; +/* don't alter vm_start if the coredump is running */ +if (!mmget_still_valid (mm)) +return NULL ; start = vma->vm_start;if (expand_stack (vma, addr))return NULL ;
Signature · 1 line
2.20.1
PATCH v2 1/1 coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping
From: Andrea Arcangeli <[email protected]>
Upstream 04f5866e41fb70690e28397487d8bd8eea7d712a commit.
The core dumping code has always run without holding the mmap_sem for writing, despite that is the only way to ensure that the entire vma layout will not change from under it. Only using some signal serialization on the processes belonging to the mm is not nearly enough. This was pointed out earlier. For example in Hugh's post from Jul 2017:
https://lkml.kernel.org/r/[email protected]
"Not strictly relevant here, but a related note: I was very surprised
to discover, only quite recently, how handle_mm_fault() may be called without down_read(mmap_sem) - when core dumping. That seems a misguided optimization to me, which would also be nice to correct"
In particular because the growsdown and growsup can move the vm_start/vm_end the various loops the core dump does around the vma will not be consistent if page faults can happen concurrently.
Pretty much all users calling mmget_not_zero()/get_task_mm() and then taking the mmap_sem had the potential to introduce unexpected side effects in the core dumping code.
Adding mmap_sem for writing around the ->core_dump invocation is a viable long term fix, but it requires removing all copy user and page faults and to replace them with get_dump_page() for all binary formats which is not suitable as a short term fix.
For the time being this solution manually covers the places that can confuse the core dump either by altering the vma layout or the vma flags while it runs. Once ->core_dump runs under mmap_sem for writing the function mmget_still_valid() can be dropped.
Allowing mmap_sem protected sections to run in parallel with the coredump provides some minor parallelism advantage to the swapoff code (which seems to be safe enough by never mangling any vma field and can keep doing swapins in parallel to the core dumping) and to some other corner case.
In order to facilitate the backporting I added "Fixes: 86039bd3b4e6" however the side effect of this same race condition in /proc/pid/mem should be reproducible since before 2.6.12-rc2 so I couldn't add any other "Fixes:" because there's no hash beyond the git genesis commit.
Because find_extend_vma() is the only location outside of the process context that could modify the "mm" structures under mmap_sem for reading, by adding the mmget_still_valid() check to it, all other cases that take the mmap_sem for reading don't need the new check after mmget_not_zero()/get_task_mm(). The expand_stack() in page fault context also doesn't need the new check, because all tasks under core dumping are frozen.
[[email protected]: stable 4.4 backport
- drop infiniband part because of missing 5f9794dc94f59
- drop userfaultfd_event_wait_completion hunk because of missing 9cd75c3cd4c3d]
- handle binder_update_page_range because of missing 720c241924046
]
---
Hi,
this is v2 of the backport. It turned out that the khugepaged part of my backport was in fact missing in the upstream commit as well so Anrea has posted a follow up fix with the full explanation [1] and that patch should be routed to stable trees on its own.
A further review of the backport is highly appreciated of course.
[1] http://lkml.kernel.org/r/[email protected]
drivers/android/binder.c | 6 ++++++ fs/proc/task_mmu.c | 18 ++++++++++++++++++ fs/userfaultfd.c | 10 ++++++++-- include/linux/mm.h | 21 +++++++++++++++++++++ mm/mmap.c | 7 ++++++- 5 files changed, 59 insertions(+), 3 deletions(-)
drivers/android/binder.c +6 -0
@@ -570,6 +570,12 @@ static int binder_update_page_range(struct binder_proc *proc, int allocate,if (mm) {down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) { +if (allocate ==0 ) + goto free_range; + goto err_no_vma; + } + vma = proc->vma;if (vma && mm != proc->vma_vm_mm) { pr_err("%d: vma mm and task mm mismatch\n",
fs/proc/task_mmu.c +18 -0
@@ -947,6 +947,24 @@ static ssize_t clear_refs_write(struct file *file, const char __user *buf,continue ;up_read (&mm->mmap_sem);down_write (&mm->mmap_sem); +/* +* Avoid to modify vma->vm_flags +* without locked ops while the +* coredump reads the vm_flags. +*/ +if (!mmget_still_valid (mm)) { +/* +* Silently return "count" +* like if get_task_mm() +* failed. FIXME: should this +* function have returned +* -ESRCH if get_task_mm() +* failed like if +* get_proc_task() fails? +*/ +up_write (&mm->mmap_sem); + goto out_mm; + }for (vma = mm->mmap; vma; vma = vma->vm_next) { vma->vm_flags &= ~VM_SOFTDIRTY ;vma_set_page_prot (vma);
fs/userfaultfd.c +8 -2
@@ -443,6 +443,8 @@ static int userfaultfd_release(struct inode *inode, struct file *file)* taking the mmap_sem for writing. */ down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) + goto skip_mm; prev =NULL ;for (vma = mm->mmap; vma; vma = vma->vm_next) {cond_resched ();
@@ -465,6 +467,7 @@ static int userfaultfd_release(struct inode *inode, struct file *file) vma->vm_flags = new_flags; vma->vm_userfaultfd_ctx =NULL_VM_UFFD_CTX ; } +skip_mm:up_write (&mm->mmap_sem);/*
@@ -761,9 +764,10 @@ static int userfaultfd_register(struct userfaultfd_ctx *ctx, end = start + uffdio_register.range.len;down_write (&mm->mmap_sem); - vma =find_vma_prev (mm, start, &prev); - ret = -ENOMEM ; +if (!mmget_still_valid (mm)) + goto out_unlock; + vma =find_vma_prev (mm, start, &prev);if (!vma) goto out_unlock;
@@ -903,6 +907,8 @@ static int userfaultfd_unregister(struct userfaultfd_ctx *ctx, end = start + uffdio_unregister.len;down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) + goto out_unlock; vma =find_vma_prev (mm, start, &prev); ret = -ENOMEM ;
include/linux/mm.h +21 -0
@@ -1098,6 +1098,27 @@ void zap_page_range(struct vm_area_struct *vma, unsigned long address,void unmap_vmas (struct mmu_gather *tlb,struct vm_area_struct *start_vma,unsigned long start,unsigned long end); +/* +* This has to be called after a get_task_mm()/mmget_not_zero() +* followed by taking the mmap_sem for writing before modifying the +* vmas or anything the coredump pretends not to change from under it. +* +* NOTE: find_extend_vma() called from GUP context is the only place +* that can modify the "mm" (notably the vm_start/end) under mmap_sem +* for reading and outside the context of the process, so it is also +* the only case that holds the mmap_sem for reading that must call +* this function. Generally if the mmap_sem is hold for reading +* there's no need of this check after get_task_mm()/mmget_not_zero(). +* +* This function can be obsoleted and the check can be removed, after +* the coredump code will hold the mmap_sem for writing before +* invoking the ->core_dump methods. +*/ +static inline bool mmget_still_valid (struct mm_struct *mm) +{ +return likely (!mm->core_state); +} +/** * mm_walk - callbacks for walk_page_range * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
mm/mmap.c +6 -1
@@ -42,6 +42,7 @@#include <linux/memory.h> #include <linux/printk.h> #include <linux/userfaultfd_k.h> +#include <linux/mm.h> #include <asm/uaccess.h> #include <asm/cacheflush.h>
@@ -2398,7 +2399,8 @@ find_extend_vma(struct mm_struct *mm, unsigned long addr) vma =find_vma_prev (mm, addr, &prev);if (vma && (vma->vm_start <= addr))return vma; -if (!prev ||expand_stack (prev, addr)) +/* don't alter vm_end if the coredump is running */ +if (!prev || !mmget_still_valid (mm) ||expand_stack (prev, addr))return NULL ;if (prev->vm_flags &VM_LOCKED )populate_vma_page_range (prev, addr, prev->vm_end,NULL );
@@ -2424,6 +2426,9 @@ find_extend_vma(struct mm_struct *mm, unsigned long addr)return vma;if (!(vma->vm_flags &VM_GROWSDOWN ))return NULL ; +/* don't alter vm_start if the coredump is running */ +if (!mmget_still_valid (mm)) +return NULL ; start = vma->vm_start;if (expand_stack (vma, addr))return NULL ;
Signature · 1 line
2.20.1
Just a heads up. Ajay Kaher has noticed that mlx4 driver is missing the check in 4.14 [1] and 4.4 seems to have the same problem. I will wait for more review before reposting v3. The incremental diff is:
drivers/infiniband/hw/mlx4/main.c +3 -0
@@ -1042,6 +1042,8 @@ static void mlx4_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)* mlx4_ib_vma_close(). */ down_write (&owning_mm->mmap_sem); +if (!mmget_still_valid (owning_mm)) + goto skip_mm;for (i =0 ; i <HW_BAR_COUNT ; i++) { vma = context->hw_bar_info[i].vma;if (!vma)
@@ -1061,6 +1063,7 @@ static void mlx4_ib_disassociate_ucontext(struct ib_ucontext *ibcontext) context->hw_bar_info[i].vma->vm_ops =NULL ; } +skip_mm:up_write (&owning_mm->mmap_sem);mmput (owning_mm);put_task_struct (owning_process);
Signature · 2 lines
Michal Hocko SUSE Labs
Quoted 17 lines · Michal Hocko
Just a heads up. Ajay Kaher has noticed that mlx4 driver is missing the check in 4.14 [1] and 4.4 seems to have the same problem. I will wait for more review before reposting v3. The incremental diff is:
drivers/infiniband/hw/mlx4/main.c +3
-0+++ b/drivers/infiniband/hw/mlx4/main.c@@ -1042,6 +1042,8 @@ static void mlx4_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)* mlx4_ib_vma_close(). */ down_write (&owning_mm->mmap_sem); +if (!mmget_still_valid (owning_mm)) + goto skip_mm;for (i =0 ; i <HW_BAR_COUNT ; i++) { vma = context->hw_bar_info[i].vma;if (!vma)
drivers/infiniband/hw/mlx4/main.c static void mlx4_ib_disassociate_ucontext(struct ib_ucontext *ibcontext) · line 1069
Quoted 7 lines
drivers/infiniband/hw/mlx4/main.c +1
-0@@ -1061,6 +1063,7 @@ static void mlx4_ib_disassociate_ucontext(struct ib_ucontext *ibcontext) context->hw_bar_info[i].vma->vm_ops =NULL ; } +skip_mm:up_write (&owning_mm->mmap_sem);mmput (owning_mm);put_task_struct (owning_process);
Looks OK to me
Jason
PATCH v3 1/1 coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping
In reply to Michal HockoFrom: Andrea Arcangeli <[email protected]>
Upstream 04f5866e41fb70690e28397487d8bd8eea7d712a commit.
The core dumping code has always run without holding the mmap_sem for writing, despite that is the only way to ensure that the entire vma layout will not change from under it. Only using some signal serialization on the processes belonging to the mm is not nearly enough. This was pointed out earlier. For example in Hugh's post from Jul 2017:
https://lkml.kernel.org/r/[email protected]
"Not strictly relevant here, but a related note: I was very surprised
to discover, only quite recently, how handle_mm_fault() may be called without down_read(mmap_sem) - when core dumping. That seems a misguided optimization to me, which would also be nice to correct"
In particular because the growsdown and growsup can move the vm_start/vm_end the various loops the core dump does around the vma will not be consistent if page faults can happen concurrently.
Pretty much all users calling mmget_not_zero()/get_task_mm() and then taking the mmap_sem had the potential to introduce unexpected side effects in the core dumping code.
Adding mmap_sem for writing around the ->core_dump invocation is a viable long term fix, but it requires removing all copy user and page faults and to replace them with get_dump_page() for all binary formats which is not suitable as a short term fix.
For the time being this solution manually covers the places that can confuse the core dump either by altering the vma layout or the vma flags while it runs. Once ->core_dump runs under mmap_sem for writing the function mmget_still_valid() can be dropped.
Allowing mmap_sem protected sections to run in parallel with the coredump provides some minor parallelism advantage to the swapoff code (which seems to be safe enough by never mangling any vma field and can keep doing swapins in parallel to the core dumping) and to some other corner case.
In order to facilitate the backporting I added "Fixes: 86039bd3b4e6" however the side effect of this same race condition in /proc/pid/mem should be reproducible since before 2.6.12-rc2 so I couldn't add any other "Fixes:" because there's no hash beyond the git genesis commit.
Because find_extend_vma() is the only location outside of the process context that could modify the "mm" structures under mmap_sem for reading, by adding the mmget_still_valid() check to it, all other cases that take the mmap_sem for reading don't need the new check after mmget_not_zero()/get_task_mm(). The expand_stack() in page fault context also doesn't need the new check, because all tasks under core dumping are frozen.
[[email protected]: stable 4.4 backport
- drop infiniband part because of missing 5f9794dc94f59
- drop userfaultfd_event_wait_completion hunk because of missing 9cd75c3cd4c3d]
- handle binder_update_page_range because of missing 720c241924046
- handle mlx5_ib_disassociate_ucontext - [email protected]
]
--- drivers/android/binder.c | 6 ++++++ drivers/infiniband/hw/mlx4/main.c | 3 +++ fs/proc/task_mmu.c | 18 ++++++++++++++++++ fs/userfaultfd.c | 10 ++++++++-- include/linux/mm.h | 21 +++++++++++++++++++++ mm/mmap.c | 7 ++++++- 6 files changed, 62 insertions(+), 3 deletions(-)
drivers/android/binder.c +6 -0
@@ -570,6 +570,12 @@ static int binder_update_page_range(struct binder_proc *proc, int allocate,if (mm) {down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) { +if (allocate ==0 ) + goto free_range; + goto err_no_vma; + } + vma = proc->vma;if (vma && mm != proc->vma_vm_mm) { pr_err("%d: vma mm and task mm mismatch\n",
drivers/infiniband/hw/mlx4/main.c +3 -0
@@ -1042,6 +1042,8 @@ static void mlx4_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)* mlx4_ib_vma_close(). */ down_write (&owning_mm->mmap_sem); +if (!mmget_still_valid (owning_mm)) + goto skip_mm;for (i =0 ; i <HW_BAR_COUNT ; i++) { vma = context->hw_bar_info[i].vma;if (!vma)
@@ -1061,6 +1063,7 @@ static void mlx4_ib_disassociate_ucontext(struct ib_ucontext *ibcontext) context->hw_bar_info[i].vma->vm_ops =NULL ; } +skip_mm:up_write (&owning_mm->mmap_sem);mmput (owning_mm);put_task_struct (owning_process);
fs/proc/task_mmu.c +18 -0
@@ -947,6 +947,24 @@ static ssize_t clear_refs_write(struct file *file, const char __user *buf,continue ;up_read (&mm->mmap_sem);down_write (&mm->mmap_sem); +/* +* Avoid to modify vma->vm_flags +* without locked ops while the +* coredump reads the vm_flags. +*/ +if (!mmget_still_valid (mm)) { +/* +* Silently return "count" +* like if get_task_mm() +* failed. FIXME: should this +* function have returned +* -ESRCH if get_task_mm() +* failed like if +* get_proc_task() fails? +*/ +up_write (&mm->mmap_sem); + goto out_mm; + }for (vma = mm->mmap; vma; vma = vma->vm_next) { vma->vm_flags &= ~VM_SOFTDIRTY ;vma_set_page_prot (vma);
fs/userfaultfd.c +8 -2
@@ -443,6 +443,8 @@ static int userfaultfd_release(struct inode *inode, struct file *file)* taking the mmap_sem for writing. */ down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) + goto skip_mm; prev =NULL ;for (vma = mm->mmap; vma; vma = vma->vm_next) {cond_resched ();
@@ -465,6 +467,7 @@ static int userfaultfd_release(struct inode *inode, struct file *file) vma->vm_flags = new_flags; vma->vm_userfaultfd_ctx =NULL_VM_UFFD_CTX ; } +skip_mm:up_write (&mm->mmap_sem);/*
@@ -761,9 +764,10 @@ static int userfaultfd_register(struct userfaultfd_ctx *ctx, end = start + uffdio_register.range.len;down_write (&mm->mmap_sem); - vma =find_vma_prev (mm, start, &prev); - ret = -ENOMEM ; +if (!mmget_still_valid (mm)) + goto out_unlock; + vma =find_vma_prev (mm, start, &prev);if (!vma) goto out_unlock;
@@ -903,6 +907,8 @@ static int userfaultfd_unregister(struct userfaultfd_ctx *ctx, end = start + uffdio_unregister.len;down_write (&mm->mmap_sem); +if (!mmget_still_valid (mm)) + goto out_unlock; vma =find_vma_prev (mm, start, &prev); ret = -ENOMEM ;
include/linux/mm.h +21 -0
@@ -1098,6 +1098,27 @@ void zap_page_range(struct vm_area_struct *vma, unsigned long address,void unmap_vmas (struct mmu_gather *tlb,struct vm_area_struct *start_vma,unsigned long start,unsigned long end); +/* +* This has to be called after a get_task_mm()/mmget_not_zero() +* followed by taking the mmap_sem for writing before modifying the +* vmas or anything the coredump pretends not to change from under it. +* +* NOTE: find_extend_vma() called from GUP context is the only place +* that can modify the "mm" (notably the vm_start/end) under mmap_sem +* for reading and outside the context of the process, so it is also +* the only case that holds the mmap_sem for reading that must call +* this function. Generally if the mmap_sem is hold for reading +* there's no need of this check after get_task_mm()/mmget_not_zero(). +* +* This function can be obsoleted and the check can be removed, after +* the coredump code will hold the mmap_sem for writing before +* invoking the ->core_dump methods. +*/ +static inline bool mmget_still_valid (struct mm_struct *mm) +{ +return likely (!mm->core_state); +} +/** * mm_walk - callbacks for walk_page_range * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
mm/mmap.c +6 -1
@@ -42,6 +42,7 @@#include <linux/memory.h> #include <linux/printk.h> #include <linux/userfaultfd_k.h> +#include <linux/mm.h> #include <asm/uaccess.h> #include <asm/cacheflush.h>
@@ -2398,7 +2399,8 @@ find_extend_vma(struct mm_struct *mm, unsigned long addr) vma =find_vma_prev (mm, addr, &prev);if (vma && (vma->vm_start <= addr))return vma; -if (!prev ||expand_stack (prev, addr)) +/* don't alter vm_end if the coredump is running */ +if (!prev || !mmget_still_valid (mm) ||expand_stack (prev, addr))return NULL ;if (prev->vm_flags &VM_LOCKED )populate_vma_page_range (prev, addr, prev->vm_end,NULL );
@@ -2424,6 +2426,9 @@ find_extend_vma(struct mm_struct *mm, unsigned long addr)return vma;if (!(vma->vm_flags &VM_GROWSDOWN ))return NULL ; +/* don't alter vm_start if the coredump is running */ +if (!mmget_still_valid (mm)) +return NULL ; start = vma->vm_start;if (expand_stack (vma, addr))return NULL ;
Signature · 1 line
2.20.1
Greg KH
maintainerQuoted 77 lines · Michal Hocko
From: Andrea Arcangeli <[email protected]>
Upstream
04f5866e41fb70690e28397487d8bd8eea7d712acommit.The core dumping code has always run without holding the
mmap_semfor writing, despite that is the only way to ensure that the entire vma layout will not change from under it. Only using some signal serialization on the processes belonging to the mm is not nearly enough. This was pointed out earlier. For example in Hugh's post from Jul 2017:https://lkml.kernel.org/r/[email protected]
"Not strictly relevant here, but a related note: I was very surprised
to discover, only quite recently, howhandle_mm_fault()may be called withoutdown_read(mmap_sem)- when core dumping. That seems a misguided optimization to me, which would also be nice to correct"In particular because the growsdown and growsup can move the vm_start/vm_end the various loops the core dump does around the vma will not be consistent if page faults can happen concurrently.
Pretty much all users calling mmget_not_zero()/get_task_mm() and then taking the
mmap_semhad the potential to introduce unexpected side effects in the core dumping code.Adding
mmap_semfor writing around the ->core_dump invocation is a viable long term fix, but it requires removing all copy user and page faults and to replace them withget_dump_page()for all binary formats which is not suitable as a short term fix.For the time being this solution manually covers the places that can confuse the core dump either by altering the vma layout or the vma flags while it runs. Once ->core_dump runs under
mmap_semfor writing the functionmmget_still_valid()can be dropped.Allowing
mmap_semprotected sections to run in parallel with the coredump provides some minor parallelism advantage to the swapoff code (which seems to be safe enough by never mangling any vma field and can keep doing swapins in parallel to the core dumping) and to some other corner case.In order to facilitate the backporting I added "Fixes:
86039bd3b4e6" however the side effect of this same race condition in/proc/pid/memshould be reproducible since before 2.6.12-rc2 so I couldn't add any other "Fixes:" because there's no hash beyond the git genesis commit.Because
find_extend_vma()is the only location outside of the process context that could modify the "mm" structures undermmap_semfor reading, by adding themmget_still_valid()check to it, all other cases that take themmap_semfor reading don't need the new check after mmget_not_zero()/get_task_mm(). Theexpand_stack()in page fault context also doesn't need the new check, because all tasks under core dumping are frozen.Fixes 86039bd3b4e6 ("userfaultfd: add new syscall to provide memory externalization")Signed-off-by Andrea Arcangeli <[email protected]>Reported-by Jann Horn <[email protected]>Suggested-by Oleg Nesterov <[email protected]>Acked-by Peter Xu <[email protected]>Reviewed-by Mike Rapoport <[email protected]>Reviewed-by Oleg Nesterov <[email protected]>Reviewed-by Jann Horn <[email protected]>Acked-by Jason Gunthorpe <[email protected]>Acked-by Michal Hocko <[email protected]>Cc <[email protected]>Signed-off-by Andrew Morton <[email protected]>Signed-off-by Linus Torvalds <[email protected]>Cc Joel Fernandes (Google) <[email protected]>[[email protected]: stable 4.4 backport
- drop infiniband part because of missing5f9794dc94f59
- dropuserfaultfd_event_wait_completionhunk because of missing9cd75c3cd4c3d]
- handlebinder_update_page_rangebecause of missing720c241924046
- handlemlx5_ib_disassociate_ucontext- [email protected]
]Signed-off-by Michal Hocko <[email protected]>
Quoted 8 lines
--- drivers/android/binder.c | 6 ++++++ drivers/infiniband/hw/mlx4/main.c | 3 +++ fs/proc/task_mmu.c | 18 ++++++++++++++++++ fs/userfaultfd.c | 10 ++++++++-- include/linux/mm.h | 21 +++++++++++++++++++++ mm/mmap.c | 7 ++++++- 6 files changed, 62 insertions(+), 3 deletions(-)
I've queued this up now, as it looks like everyone agrees with it. What about a 4.9.y backport?
thanks,
greg k-h