shmem: fix double uncharge in __shmem_file_setup()
Subsystems (3)
PATCH 1/3 shmem: fix double uncharge in __shmem_file_setup()
If __shmem_file_setup() fails on struct file allocation it uncharges memory commitment twice: first by shmem_unacct_size() and second time implicitly in shmem_evict_inode() when it kills newly created inode.
This patch removes shmem_unacct_size() from error path if inode already here.
--- mm/shmem.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
mm/shmem.c +6 -6
@@ -2895,16 +2895,16 @@ static struct file *__shmem_file_setup(const char *name, loff_t size, this.len =strlen (name); this.hash =0 ;/* will go */ sb = shm_mnt->mnt_sb; + path.mnt =mntget (shm_mnt); path.dentry =d_alloc_pseudo (sb, &this);if (!path.dentry) goto put_memory;d_set_d_op (path.dentry, &anon_ops); - path.mnt =mntget (shm_mnt); res =ERR_PTR (-ENOSPC ); inode =shmem_get_inode (sb,NULL ,S_IFREG |S_IRWXUGO ,0 , flags);if (!inode) - goto put_dentry; + goto put_memory; inode->i_flags |= i_flags;d_instantiate (path.dentry, inode);
@@ -2912,19 +2912,19 @@ static struct file *__shmem_file_setup(const char *name, loff_t size,clear_nlink (inode);/* It is unlinked */ res =ERR_PTR (ramfs_nommu_expand_for_mapping (inode, size));if (IS_ERR (res)) - goto put_dentry; + goto put_path; res =alloc_file (&path,FMODE_WRITE |FMODE_READ , &shmem_file_operations);if (IS_ERR (res)) - goto put_dentry; + goto put_path;return res; -put_dentry: -path_put (&path); put_memory:shmem_unacct_size (flags, size); +put_path: +path_put (&path);return res; }
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
PATCH 2/3 shmem: update memory reservation on truncate
Shared anonymous mapping created without MAP_NORESERVE holds memory reservation for whole range of shmem segment. Usually there is no way to change its size, but /proc/<pid>/map_files/...
(available if CONFIG_CHECKPOINT_RESTORE=y) allows to do that.
This patch adjust memory reservation in shmem_setattr().
---
exploit:
#include <sys/mman.h>
#include <unistd.h>
#include <stdio.h>
int main(int argc, char **argv)
{
unsigned long addr;
char path[100];
/* charge 4KiB */
addr = (unsigned long)mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED|MAP_ANONYMOUS, -1, 0);
sprintf(path, "/proc/self/map_files/%lx-%lx", addr, addr + 4096);
truncate(path, 1 << 30);
/* uncharge 1GiB */
}
---
mm/shmem.c | 17 +++++++++++++++++
1 file changed, 17 insertions(+)mm/shmem.c +17 -0
@@ -149,6 +149,19 @@ static inline void shmem_unacct_size(unsigned long flags, loff_t size)vm_unacct_memory (VM_ACCT (size)); } +static inline int shmem_reacct_size (unsigned long flags, +loff_t oldsize,loff_t newsize) +{ +if (!(flags &VM_NORESERVE )) { +if (VM_ACCT (newsize) >VM_ACCT (oldsize)) +return security_vm_enough_memory_mm (current->mm, +VM_ACCT (newsize) -VM_ACCT (oldsize)); +else if (VM_ACCT (newsize) <VM_ACCT (oldsize)) +vm_unacct_memory (VM_ACCT (oldsize) -VM_ACCT (newsize)); + } +return 0 ; +} +/* * ... whereas tmpfs objects are accounted incrementally as * pages are allocated, in order to allow huge sparse files.
@@ -543,6 +556,10 @@ static int shmem_setattr(struct dentry *dentry, struct iattr *attr)loff_t newsize = attr->ia_size;if (newsize != oldsize) { + error =shmem_reacct_size (SHMEM_I (inode)->flags, + oldsize, newsize); +if (error) +return error;i_size_write (inode, newsize); inode->i_ctime = inode->i_mtime =CURRENT_TIME ; }
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
PATCH 3/3 mm: catch memory commitment underflow
In reply to Konstantin KhlebnikovThis patch prints warning (if CONFIG_DEBUG_VM=y) when memory commitment becomes too negative.
--- mm/mmap.c | 6 ++++++ 1 file changed, 6 insertions(+)
mm/mmap.c +6 -0
@@ -134,6 +134,12 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) {unsigned long free, allowed, reserve; +#ifdef CONFIG_DEBUG_VM +WARN_ONCE (percpu_counter_read (&vm_committed_as) < + -(s64 )vm_committed_as_batch *num_online_cpus (), +"memory commitment underflow" ); +#endif +vm_acct_memory (pages);/*
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Andrew Morton
maintainerRe: [PATCH 3/3] mm: catch memory commitment underflow
Quoted 8 lines · On Wed, 25 Jun 2014 00:16:14 +0400 Konstantin Khlebnikov <[email protected]>…
This patch prints warning (if
CONFIG_DEBUG_VM=y) when memory commitment becomes too negative....
mm/mmap.c int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) · line 143
Quoted 10 lines
mm/mmap.c +6
-0@@ -134,6 +134,12 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) {unsigned long free, allowed, reserve; +#ifdef CONFIG_DEBUG_VM +WARN_ONCE (percpu_counter_read (&vm_committed_as) < + -(s64 )vm_committed_as_batch *num_online_cpus (), +"memory commitment underflow" ); +#endif +vm_acct_memory (pages);
The changelog doesn't describe the reasons for making the change.
I assume this warning will detect the situation which the previous two patches just fixed?
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm: catch memory commitment underflow
In reply to Konstantin Khlebnikovmm/mmap.c int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) · line 141
Quoted 7 lines · Konstantin Khlebnikov
reserve;
mm/mmap.c +5
-0+#ifdef CONFIG_DEBUG_VM +WARN_ONCE (percpu_counter_read (&vm_committed_as) < + -(s64 )vm_committed_as_batch *num_online_cpus (), +"memory commitment underflow" ); +#endif
Why not use VM_WARN_ON_ONCE()?
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 6 lines · Konstantin Khlebnikov
If
__shmem_file_setup()fails on struct file allocation it uncharges memory commitment twice: first byshmem_unacct_size()and second time implicitly inshmem_evict_inode()when it kills newly created inode.
This patch removesshmem_unacct_size()from error path if inode already here.Signed-off-by Konstantin Khlebnikov <[email protected]>
Thank you for the patch, and thank you for your patience (or perhaps for your kindly concealed impatience): I realize that this (and the other two) have been languishing in the must-get-to-look-at-it-sometime end of my mailbox for nine months now - sorry.
Quoted 41 lines
--- mm/shmem.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)mm/shmem.c +4
-4@@ -2895,16 +2895,16 @@ static struct file *__shmem_file_setup(const char *name, loff_t size, this.len =strlen (name); this.hash =0 ;/* will go */ sb = shm_mnt->mnt_sb; + path.mnt =mntget (shm_mnt); path.dentry =d_alloc_pseudo (sb, &this);if (!path.dentry) goto put_memory;d_set_d_op (path.dentry, &anon_ops); - path.mnt =mntget (shm_mnt); res =ERR_PTR (-ENOSPC ); inode =shmem_get_inode (sb,NULL ,S_IFREG |S_IRWXUGO ,0 , flags);if (!inode) - goto put_dentry; + goto put_memory; inode->i_flags |= i_flags;d_instantiate (path.dentry, inode);@@ -2912,19 +2912,19 @@ static struct file *__shmem_file_setup(const char *name, loff_t size,clear_nlink (inode);/* It is unlinked */ res =ERR_PTR (ramfs_nommu_expand_for_mapping (inode, size));if (IS_ERR (res)) - goto put_dentry; + goto put_path; res =alloc_file (&path,FMODE_WRITE |FMODE_READ , &shmem_file_operations);if (IS_ERR (res)) - goto put_dentry; + goto put_path;return res;
mm/shmem.c static struct file *__shmem_file_setup(const char *name, loff_t size, · line 2930
Quoted 10 lines
mm/shmem.c +2
-2-put_dentry: -path_put (&path); put_memory:shmem_unacct_size (flags, size); +put_path: +path_put (&path);return res; }
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] shmem: update memory reservation on truncate
In reply to Konstantin KhlebnikovQuoted 8 lines · Konstantin Khlebnikov
Shared anonymous mapping created without
MAP_NORESERVEholds memory reservation for whole range of shmem segment. Usually there is no way to change its size, but /proc/<pid>/map_files/...
(available ifCONFIG_CHECKPOINT_RESTORE=y) allows to do that.This patch adjust memory reservation in
shmem_setattr().Signed-off-by Konstantin Khlebnikov <[email protected]>
Thank you, I knew nothing about this backdoor to shmem objects. Scary. Was this really the only problem map_files access leads to? If you did not do so already, please try to think through other possibilities.
I haven't begun, but perhaps it's not so bad. I guess the interaction with mremap extension is benign - it's annoyed people in the past that the underlying shmem object is not extended, but now here's a way that it can be.
(I'll leave it to others comment on 3/3 if they wish.)
Quoted 51 lines
---
exploit:
#include <
sys/mman.h>
#include <unistd.h>
#include <stdio.h>int
main(int argc, char **argv){ unsigned long addr; char path[100]; /* charge 4KiB */ addr = (unsigned long)mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED|MAP_ANONYMOUS, -1, 0); sprintf(path, "/proc/self/map_files/%lx-%lx", addr, addr + 4096); truncate(path, 1 << 30); /* uncharge 1GiB */ } --- mm/shmem.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)mm/shmem.c +13
-0@@ -149,6 +149,19 @@ static inline void shmem_unacct_size(unsigned long flags, loff_t size)vm_unacct_memory (VM_ACCT (size)); } +static inline int shmem_reacct_size (unsigned long flags, +loff_t oldsize,loff_t newsize) +{ +if (!(flags &VM_NORESERVE )) { +if (VM_ACCT (newsize) >VM_ACCT (oldsize)) +return security_vm_enough_memory_mm (current->mm, +VM_ACCT (newsize) -VM_ACCT (oldsize)); +else if (VM_ACCT (newsize) <VM_ACCT (oldsize)) +vm_unacct_memory (VM_ACCT (oldsize) -VM_ACCT (newsize)); + } +return 0 ; +} +/* * ... whereas tmpfs objects are accounted incrementally as * pages are allocated, in order to allow huge sparse files.
mm/shmem.c static int shmem_setattr(struct dentry *dentry, struct iattr *attr) · line 565
Quoted 10 lines
mm/shmem.c +4
-0@@ -543,6 +556,10 @@ static int shmem_setattr(struct dentry *dentry, struct iattr *attr)loff_t newsize = attr->ia_size;if (newsize != oldsize) { + error =shmem_reacct_size (SHMEM_I (inode)->flags, + oldsize, newsize); +if (error) +return error;i_size_write (inode, newsize); inode->i_ctime = inode->i_mtime =CURRENT_TIME ; }
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm: catch memory commitment underflow
In reply to Andrew MortonQuoted 23 lines · Andrew Morton
Quoted 8 lines · On Wed, 25 Jun 2014 00:16:14 +0400 Konstantin Khlebnikov <[email protected]>…
This patch prints warning (if
CONFIG_DEBUG_VM=y) when memory commitment becomes too negative....
mm/mmap.c int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) · line 143
Quoted 10 lines
mm/mmap.c +6
-0@@ -134,6 +134,12 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) {unsigned long free, allowed, reserve; +#ifdef CONFIG_DEBUG_VM +WARN_ONCE (percpu_counter_read (&vm_committed_as) < + -(s64 )vm_committed_as_batch *num_online_cpus (), +"memory commitment underflow" ); +#endif +vm_acct_memory (pages);The changelog doesn't describe the reasons for making the change.
Quoted 2 lines
I assume this warning will detect the situation which the previous two patches just fixed?
Yep. Otherwise there is no way to validate these bugs, /proc/meminfo hides negative values.
Quoted 1 line
Why not use
VM_WARN_ON_ONCE()?
This patch is older than this macro.
Previously I've sent it in the last september and it was ignored. Now I've found it again in my backlog.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] shmem: update memory reservation on truncate
In reply to Hugh DickinsQuoted 9 lines · Hugh Dickins
Quoted 7 lines · Konstantin Khlebnikov
Shared anonymous mapping created without
MAP_NORESERVEholds memory reservation for whole range of shmem segment. Usually there is no way to change its size, but /proc/<pid>/map_files/...
(available ifCONFIG_CHECKPOINT_RESTORE=y) allows to do that.This patch adjust memory reservation in
shmem_setattr().
Quoted 7 lines
Quoted 1 line
Signed-off-by Konstantin Khlebnikov <[email protected]>Acked-by Hugh Dickins <[email protected]>Thank you, I knew nothing about this backdoor to shmem objects. Scary. Was this really the only problem
map_filesaccess leads to? If you did not do so already, please try to think through other possibilities.
Ouch, it's still broken. I've fixed only truncate. write_begin/write_end and fallocate might change i_size too.
Quoted 59 lines
I haven't begun, but perhaps it's not so bad. I guess the interaction with mremap extension is benign - it's annoyed people in the past that the underlying shmem object is not extended, but now here's a way that it can be.
(I'll leave it to others comment on 3/3 if they wish.)
Quoted 51 lines
---
exploit:
#include <
sys/mman.h>
#include <unistd.h>
#include <stdio.h>int
main(int argc, char **argv){ unsigned long addr; char path[100]; /* charge 4KiB */ addr = (unsigned long)mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED|MAP_ANONYMOUS, -1, 0); sprintf(path, "/proc/self/map_files/%lx-%lx", addr, addr + 4096); truncate(path, 1 << 30); /* uncharge 1GiB */ } --- mm/shmem.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)mm/shmem.c +13
-0@@ -149,6 +149,19 @@ static inline void shmem_unacct_size(unsigned long flags, loff_t size)vm_unacct_memory (VM_ACCT (size)); } +static inline int shmem_reacct_size (unsigned long flags, +loff_t oldsize,loff_t newsize) +{ +if (!(flags &VM_NORESERVE )) { +if (VM_ACCT (newsize) >VM_ACCT (oldsize)) +return security_vm_enough_memory_mm (current->mm, +VM_ACCT (newsize) -VM_ACCT (oldsize)); +else if (VM_ACCT (newsize) <VM_ACCT (oldsize)) +vm_unacct_memory (VM_ACCT (oldsize) -VM_ACCT (newsize)); + } +return 0 ; +} +/* * ... whereas tmpfs objects are accounted incrementally as * pages are allocated, in order to allow huge sparse files. @@ -543,6 +556,10 @@ static int shmem_setattr(struct dentry *dentry, struct iattr *attr)loff_t newsize = attr->ia_size;
Quoted 10 lines
mm/shmem.c static int shmem_setattr(struct dentry *dentry, struct iattr *attr) · line 565
Quoted 10 lines
mm/shmem.c +4
-0if (newsize != oldsize) { + error =shmem_reacct_size (SHMEM_I (inode)->flags, + oldsize, newsize); +if (error) +return error;i_size_write (inode, newsize); inode->i_ctime = inode->i_mtime =CURRENT_TIME ; }
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm: catch memory commitment underflow
In reply to Konstantin KhlebnikovQuoted 4 lines · Konstantin Khlebnikov
This patch prints warning (if
CONFIG_DEBUG_VM=y) when memory commitment becomes too negative.Signed-off-by Konstantin Khlebnikov <[email protected]>
Hi Konstantin,
I seem to be hitting this warning when fuzzing on the latest -next kernel:
[ 683.674323] ------------[ cut here ]------------ [ 683.675552] WARNING: CPU: 12 PID: 25654 at mm/mmap.c:157 __vm_enough_memory+0x1b7/0x1d0() [ 683.676972] memory commitment underflow [ 683.678212] Modules linked in: [ 683.678219] CPU: 12 PID: 25654 Comm: trinity-c373 Not tainted 3.19.0-rc4-next-20150116-sasha-00054-g4ad498c-dirty #1744 [ 683.678227] ffffffff9c6f7e73 ffff8802c0883a58 ffffffff9b439fb2 0000000000000000 [ 683.678231] ffff8802c0883aa8 ffff8802c0883a98 ffffffff98159e1a ffff8802c0883ae8 [ 683.678236] 0000000000000001 0000000000000000 ffffffffffff76f1 ffff8802a9749000 [ 683.678243] Call Trace: [ 683.678288] dump_stack (lib/dump_stack.c:52) [ 683.678297] warn_slowpath_common (kernel/panic.c:447) [ 683.678302] warn_slowpath_fmt (kernel/panic.c:461) [ 683.678307] __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 683.678317] cap_vm_enough_memory (security/commoncap.c:958) [ 683.678323] security_vm_enough_memory_mm (security/security.c:212) [ 683.678331] shmem_getpage_gfp (mm/shmem.c:1161) [ 683.678337] shmem_write_begin (mm/shmem.c:1495) [ 683.678343] generic_perform_write (mm/filemap.c:2491) [ 683.678447] __generic_file_write_iter (mm/filemap.c:2632) [ 683.678452] generic_file_write_iter (mm/filemap.c:2659) [ 683.678458] do_iter_readv_writev (fs/read_write.c:680) [ 683.678461] do_readv_writev (fs/read_write.c:848) [ 683.678512] vfs_writev (fs/read_write.c:893) [ 683.678515] SyS_writev (fs/read_write.c:926 fs/read_write.c:917) [ 683.678520] tracesys_phase2 (arch/x86/kernel/entry_64.S:529)
Thanks,
Sasha
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm: catch memory commitment underflow
Quoted 9 lines · Sasha Levin
Quoted 4 lines · Konstantin Khlebnikov
This patch prints warning (if
CONFIG_DEBUG_VM=y) when memory commitment becomes too negative.Signed-off-by Konstantin Khlebnikov <[email protected]>Hi Konstantin,
I seem to be hitting this warning when fuzzing on the latest -next kernel:
That might be unexpected change of shmem file which holds anon-vma data, thanks to checkpoint-restore they are expoted via /proc/.../map_files
I've fixed truncate (https://lkml.org/lkml/2014/6/24/729) but there are some other ways
to change i_size: write, fallocate and maybe something else.
We could seal this promblem (literally).
mm/shmem.c +0 -0
@@ -3355,6 +3355,9 @@ static struct file *__shmem_file_setup(const
char *name, loff_t size,
if (!inode)
goto put_memory;+ if (!(flags & VM_NORESERVE))
+ SHMEM_I(inode)->seals |= F_SEAL_SHRINK | F_SEAL_GROW;
+
inode->i_flags |= i_flags;
d_instantiate(path.dentry, inode);
inode->i_size = size;Quoted 28 lines
[ 683.674323] ------------[ cut here ]------------ [ 683.675552] WARNING: CPU: 12 PID: 25654 at mm/mmap.c:157 __vm_enough_memory+0x1b7/0x1d0() [ 683.676972] memory commitment underflow [ 683.678212] Modules linked in: [ 683.678219] CPU: 12 PID: 25654 Comm: trinity-c373 Not tainted 3.19.0-rc4-next-20150116-sasha-00054-g4ad498c-dirty #1744 [ 683.678227] ffffffff9c6f7e73 ffff8802c0883a58 ffffffff9b439fb2 0000000000000000 [ 683.678231] ffff8802c0883aa8 ffff8802c0883a98 ffffffff98159e1a ffff8802c0883ae8 [ 683.678236] 0000000000000001 0000000000000000 ffffffffffff76f1 ffff8802a9749000 [ 683.678243] Call Trace: [ 683.678288] dump_stack (lib/dump_stack.c:52) [ 683.678297] warn_slowpath_common (kernel/panic.c:447) [ 683.678302] warn_slowpath_fmt (kernel/panic.c:461) [ 683.678307] __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 683.678317] cap_vm_enough_memory (security/commoncap.c:958) [ 683.678323] security_vm_enough_memory_mm (security/security.c:212) [ 683.678331] shmem_getpage_gfp (mm/shmem.c:1161) [ 683.678337] shmem_write_begin (mm/shmem.c:1495) [ 683.678343] generic_perform_write (mm/filemap.c:2491) [ 683.678447] __generic_file_write_iter (mm/filemap.c:2632) [ 683.678452] generic_file_write_iter (mm/filemap.c:2659) [ 683.678458] do_iter_readv_writev (fs/read_write.c:680) [ 683.678461] do_readv_writev (fs/read_write.c:848) [ 683.678512] vfs_writev (fs/read_write.c:893) [ 683.678515] SyS_writev (fs/read_write.c:926 fs/read_write.c:917) [ 683.678520] tracesys_phase2 (arch/x86/kernel/entry_64.S:529)
Quoted 2 lines
Thanks,
Sasha
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm: catch memory commitment underflow
Quoted 9 lines · Konstantin Khlebnikov
Quoted 8 lines · Sasha Levin
Quoted 4 lines · Konstantin Khlebnikov
This patch prints warning (if
CONFIG_DEBUG_VM=y) when memory commitment becomes too negative.Signed-off-by Konstantin Khlebnikov <[email protected]>Hi Konstantin,
Quoted 8 lines
Quoted 1 line
I seem to be hitting this warning when fuzzing on the latest -next kernel:
That might be unexpected change of shmem file which holds anon-vma data, thanks to checkpoint-restore they are expoted via
/proc/.../map_filesI've fixed truncate (https://lkml.org/lkml/2014/6/24/729) but there are some other ways
to changei_size: write, fallocate and maybe something else.
deja vu!
With the latest -next:
[ 884.898243] ------------[ cut here ]------------ [ 884.899983] ------------[ cut here ]------------ [ 884.900013] WARNING: CPU: 5 PID: 17543 at mm/mmap.c:159 __vm_enough_memory+0x3b7/0x440() [ 884.900017] ------------[ cut here ]------------ [ 884.900155] memory commitment underflow [ 884.900158] Modules linked in: [ 884.900167] CPU: 5 PID: 17543 Comm: trinity-c102 Not tainted 4.0.0-next-20150424-sasha-00038-ga61bf14 #2171 [ 884.900180] WARNING: CPU: 0 PID: 18483 at mm/mmap.c:159 __vm_enough_memory+0x3b7/0x440() [ 884.900185] ffff88017e180000 [ 884.900188] memory commitment underflow [ 884.900190] 0000000012331894 [ 884.900193] Modules linked in: [ 884.900195] ffff8807dd8bf5a8 [ 884.900196] [ 884.900200] ffffffffa9abbf32 [ 884.900211] 0000000000000000 ffff8807dd8bf628 ffff8807dd8bf5f8 ffffffff9f1f1c2a [ 884.900222] ffff8807dd8bf5d8 ffffffff9f5efb27 ffff8807dd8bf628 ffffed00fbb17ec1 [ 884.900230] Call Trace: [ 884.900247] dump_stack (lib/dump_stack.c:52) [ 884.900260] warn_slowpath_common (kernel/panic.c:447) [ 884.900270] ? __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 884.900278] warn_slowpath_fmt (kernel/panic.c:453) [ 884.900286] ? warn_slowpath_common (kernel/panic.c:453) [ 884.900300] ? find_get_entry (include/linux/rcupdate.h:969 mm/filemap.c:1003) [ 884.900310] __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 884.900317] ? find_get_entry (mm/filemap.c:967) [ 884.900334] cap_vm_enough_memory (security/commoncap.c:954) [ 884.900344] security_vm_enough_memory_mm (security/security.c:235) [ 884.900355] shmem_getpage_gfp (mm/shmem.c:1156) [ 884.900369] ? trace_hardirqs_on_thunk (arch/x86/lib/thunk_64.S:42) [ 884.900382] ? lockdep_init (include/linux/list.h:28 kernel/locking/lockdep.c:4065) [ 884.900391] ? shmem_add_to_page_cache (mm/shmem.c:1034) [ 884.900402] ? __wake_up_locked_key (kernel/sched/wait.c:456) [ 884.900414] ? __bdi_update_bandwidth (mm/page-writeback.c:1579) [ 884.900422] ? __lock_is_held (kernel/locking/lockdep.c:3572) [ 884.900432] ? iov_iter_single_seg_count (lib/iov_iter.c:310) [ 884.900441] shmem_write_begin (mm/shmem.c:1492) [ 884.900450] generic_perform_write (mm/filemap.c:2467) [ 884.900461] ? generic_write_checks (mm/filemap.c:2427) [ 884.900475] ? file_update_time (fs/inode.c:1746) [ 884.900483] ? file_remove_suid (fs/inode.c:1718) [ 884.900492] ? generic_file_write_iter (include/linux/sched.h:3091 include/linux/sched.h:3102 mm/filemap.c:2269 mm/filemap.c:2622) [ 884.900501] ? mutex_trylock (kernel/locking/mutex.c:615) [ 884.900510] __generic_file_write_iter (mm/filemap.c:2597) [ 884.900521] ? get_parent_ip (kernel/sched/core.c:2556) [ 884.900531] generic_file_write_iter (mm/filemap.c:2625) [ 884.900543] do_iter_readv_writev (fs/read_write.c:665) [ 884.900551] ? do_readv_writev (include/linux/fs.h:2417 fs/read_write.c:804) [ 884.900558] ? do_loop_readv_writev (fs/read_write.c:657) [ 884.900567] ? rw_verify_area (fs/read_write.c:406 (discriminator 4)) [ 884.900576] do_readv_writev (fs/read_write.c:808) [ 884.900583] ? __generic_file_write_iter (mm/filemap.c:2616) [ 884.900591] ? vfs_write (fs/read_write.c:777) [ 884.900601] ? debug_smp_processor_id (lib/smp_processor_id.c:57) [ 884.900609] ? get_lock_stats (kernel/locking/lockdep.c:249) [ 884.900621] ? vtime_account_user (kernel/sched/cputime.c:701) [ 884.900630] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63) [ 884.900639] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2594 kernel/locking/lockdep.c:2636) [ 884.900646] ? trace_hardirqs_on (kernel/locking/lockdep.c:2644) [ 884.900654] vfs_writev (fs/read_write.c:848) [ 884.900663] SyS_writev (fs/read_write.c:881 fs/read_write.c:872) [ 884.900671] ? SyS_readv (fs/read_write.c:872) [ 884.900684] ? syscall_trace_enter_phase2 (arch/x86/kernel/ptrace.c:1592) [ 884.900693] ? trace_hardirqs_on_thunk (arch/x86/lib/thunk_64.S:42) [ 884.900703] tracesys_phase2 (arch/x86/kernel/entry_64.S:337) [ 884.900716] ? __percpu_counter_sum (lib/percpu_counter.c:107) [ 884.900723] ---[ end trace 957b1b1a507acb40 ]--- [ 884.900730] CPU: 0 PID: 18483 Comm: trinity-c39 Not tainted 4.0.0-next-20150424-sasha-00038-ga61bf14 #2171 [ 884.900746] ffff880077ba3000 000000005dbf6d8b ffff88007bd8f5b8 ffffffffa9abbf32 [ 884.900759] 0000000000000000 ffff88007bd8f638 ffff88007bd8f608 ffffffff9f1f1c2a [ 884.900771] ffff88007bd8f618 ffffffff9f5efb27 ffffffff9f2700d0 ffffed000f7b1ec3 [ 884.900774] Call Trace: [ 884.900787] dump_stack (lib/dump_stack.c:52) [ 884.900796] warn_slowpath_common (kernel/panic.c:447) [ 884.900807] ? __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 884.900820] ? finish_task_switch (kernel/sched/sched.h:1077 kernel/sched/core.c:2245) [ 884.900830] warn_slowpath_fmt (kernel/panic.c:453) [ 884.900838] ? warn_slowpath_common (kernel/panic.c:453) [ 884.900846] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63) [ 884.900863] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2594 kernel/locking/lockdep.c:2636) [ 884.900873] __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 884.900883] ? find_get_entry (mm/filemap.c:967) [ 884.900895] cap_vm_enough_memory (security/commoncap.c:954) [ 884.900905] security_vm_enough_memory_mm (security/security.c:235) [ 884.900912] ? security_vm_enough_memory_mm (security/security.c:234) [ 884.900922] shmem_getpage_gfp (mm/shmem.c:1156) [ 884.900931] ? lockdep_init (include/linux/list.h:28 kernel/locking/lockdep.c:4065) [ 884.900940] ? shmem_add_to_page_cache (mm/shmem.c:1034) [ 884.900950] ? __wake_up_locked_key (kernel/sched/wait.c:456) [ 884.900960] ? __bdi_update_bandwidth (mm/page-writeback.c:1579) [ 884.900968] ? __lock_is_held (kernel/locking/lockdep.c:3572) [ 884.900977] ? iov_iter_single_seg_count (lib/iov_iter.c:310) [ 884.900986] shmem_write_begin (mm/shmem.c:1492) [ 884.900996] generic_perform_write (mm/filemap.c:2467) [ 884.901011] ? generic_write_checks (mm/filemap.c:2427) [ 884.901173] ? file_update_time (fs/inode.c:1746) [ 884.901182] ? file_remove_suid (fs/inode.c:1718) [ 884.901190] ? generic_file_write_iter (include/linux/sched.h:3091 include/linux/sched.h:3102 mm/filemap.c:2269 mm/filemap.c:2622) [ 884.901199] ? mutex_trylock (kernel/locking/mutex.c:615) [ 884.901207] __generic_file_write_iter (mm/filemap.c:2597) [ 884.901217] ? get_parent_ip (kernel/sched/core.c:2556) [ 884.901228] generic_file_write_iter (mm/filemap.c:2625) [ 884.901240] do_iter_readv_writev (fs/read_write.c:665) [ 884.901248] ? do_readv_writev (include/linux/fs.h:2417 fs/read_write.c:804) [ 884.901257] ? do_loop_readv_writev (fs/read_write.c:657) [ 884.901269] ? rw_verify_area (fs/read_write.c:406 (discriminator 4)) [ 884.901278] do_readv_writev (fs/read_write.c:808) [ 884.901288] ? __generic_file_write_iter (mm/filemap.c:2616) [ 884.901297] ? vfs_write (fs/read_write.c:777) [ 884.901306] ? debug_smp_processor_id (lib/smp_processor_id.c:57) [ 884.901314] ? get_lock_stats (kernel/locking/lockdep.c:249) [ 884.901326] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63) [ 884.901336] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2594 kernel/locking/lockdep.c:2636) [ 884.901345] ? trace_hardirqs_on (kernel/locking/lockdep.c:2644) [ 884.901355] vfs_writev (fs/read_write.c:848) [ 884.901365] ? __fdget_pos (fs/file.c:717) [ 884.901374] SyS_pwritev (include/linux/file.h:38 fs/read_write.c:937 fs/read_write.c:922) [ 884.901382] ? SyS_preadv (fs/read_write.c:922) [ 884.901393] ? syscall_trace_enter_phase2 (arch/x86/kernel/ptrace.c:1592) [ 884.901404] ? trace_hardirqs_on_thunk (arch/x86/lib/thunk_64.S:42) [ 884.901414] tracesys_phase2 (arch/x86/kernel/entry_64.S:337) [ 884.901423] ---[ end trace 957b1b1a507acb41 ]--- [ 885.133658] WARNING: CPU: 6 PID: 17218 at mm/mmap.c:159 __vm_enough_memory+0x3b7/0x440() [ 885.136475] memory commitment underflow [ 885.137807] Modules linked in: [ 885.139002] CPU: 6 PID: 17218 Comm: trinity-c296 Tainted: G W 4.0.0-next-20150424-sasha-00038-ga61bf14 #2171 [ 885.142889] ffff88078e4a8000 0000000057b5e6a5 ffff88078e6275a8 ffffffffa9abbf32 [ 885.145640] 0000000000000000 ffff88078e627628 ffff88078e6275f8 ffffffff9f1f1c2a [ 885.148334] ffff88078e6275d8 ffffffff9f5efb27 ffff88078e627628 ffffed00f1cc4ec1 [ 885.150427] Call Trace: [ 885.151080] dump_stack (lib/dump_stack.c:52) [ 885.152771] warn_slowpath_common (kernel/panic.c:447) [ 885.154819] ? __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 885.156674] warn_slowpath_fmt (kernel/panic.c:453) [ 885.158168] ? warn_slowpath_common (kernel/panic.c:453) [ 885.159853] ? find_get_entry (include/linux/rcupdate.h:969 mm/filemap.c:1003) [ 885.161311] __vm_enough_memory (mm/mmap.c:157 (discriminator 3)) [ 885.162823] ? find_get_entry (mm/filemap.c:967) [ 885.164599] cap_vm_enough_memory (security/commoncap.c:954) [ 885.166588] security_vm_enough_memory_mm (security/security.c:235) [ 885.168923] shmem_getpage_gfp (mm/shmem.c:1156) [ 885.170990] ? trace_hardirqs_on_thunk (arch/x86/lib/thunk_64.S:42) [ 885.173230] ? lockdep_init (include/linux/list.h:28 kernel/locking/lockdep.c:4065) [ 885.175228] ? shmem_add_to_page_cache (mm/shmem.c:1034) [ 885.177492] ? __wake_up_locked_key (kernel/sched/wait.c:456) [ 885.179648] ? __bdi_update_bandwidth (mm/page-writeback.c:1579) [ 885.181288] ? __lock_is_held (kernel/locking/lockdep.c:3572) [ 885.182750] ? __wake_up_bit (kernel/sched/wait.c:456) [ 885.184418] ? iov_iter_single_seg_count (lib/iov_iter.c:310) [ 885.186728] shmem_write_begin (mm/shmem.c:1492) [ 885.188751] generic_perform_write (mm/filemap.c:2467) [ 885.190906] ? generic_write_checks (mm/filemap.c:2427) [ 885.193252] ? file_update_time (fs/inode.c:1746) [ 885.195344] ? file_remove_suid (fs/inode.c:1718) [ 885.196996] ? generic_file_write_iter (include/linux/sched.h:3091 include/linux/sched.h:3102 mm/filemap.c:2269 mm/filemap.c:2622) [ 885.198672] ? mutex_trylock (kernel/locking/mutex.c:615) [ 885.200440] __generic_file_write_iter (mm/filemap.c:2597) [ 885.202666] ? get_parent_ip (kernel/sched/core.c:2556) [ 885.204647] generic_file_write_iter (mm/filemap.c:2625) [ 885.206875] do_iter_readv_writev (fs/read_write.c:665) [ 885.209242] ? do_readv_writev (include/linux/fs.h:2417 fs/read_write.c:804) [ 885.211306] ? do_loop_readv_writev (fs/read_write.c:657) [ 885.213414] ? rw_verify_area (fs/read_write.c:406 (discriminator 4)) [ 885.215263] do_readv_writev (fs/read_write.c:808) [ 885.216855] ? __generic_file_write_iter (mm/filemap.c:2616) [ 885.218482] ? vfs_write (fs/read_write.c:777) [ 885.219931] ? debug_smp_processor_id (lib/smp_processor_id.c:57) [ 885.222062] ? get_lock_stats (kernel/locking/lockdep.c:249) [ 885.224275] ? vtime_account_user (kernel/sched/cputime.c:701) [ 885.226375] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63) [ 885.228472] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2594 kernel/locking/lockdep.c:2636) [ 885.230673] ? trace_hardirqs_on (kernel/locking/lockdep.c:2644) [ 885.232751] vfs_writev (fs/read_write.c:848) [ 885.234490] SyS_writev (fs/read_write.c:881 fs/read_write.c:872) [ 885.236473] ? SyS_readv (fs/read_write.c:872) [ 885.238300] ? syscall_trace_enter_phase2 (arch/x86/kernel/ptrace.c:1592) [ 885.240677] ? trace_hardirqs_on_thunk (arch/x86/lib/thunk_64.S:42) [ 885.242909] tracesys_phase2 (arch/x86/kernel/entry_64.S:337) [ 885.246363] ---[ end trace 957b1b1a507acb42 ]---
Thanks,
Sasha
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm: catch memory commitment underflow
Quoted 15 lines · Sasha Levin
Quoted 14 lines · Konstantin Khlebnikov
Quoted 9 lines · Sasha Levin
Quoted 4 lines · Konstantin Khlebnikov
This patch prints warning (if
CONFIG_DEBUG_VM=y) when memory commitment becomes too negative.Signed-off-by Konstantin Khlebnikov <[email protected]>Hi Konstantin,
I seem to be hitting this warning when fuzzing on the latest -next kernel:
That might be unexpected change of shmem file which holds anon-vma data, thanks to checkpoint-restore they are expoted via
/proc/.../map_files
Quoted 7 lines
Quoted 3 lines
I've fixed truncate (https://lkml.org/lkml/2014/6/24/729) but there are some other ways
to changei_size: write, fallocate and maybe something else.deja vu!
With the latest -next:
Ping? I'm still seeing this in -next:
[668829.173774] ------------[ cut here ]------------ [668829.175259] WARNING: CPU: 18 PID: 4414 at mm/mmap.c:160 __vm_enough_memory+0x39f/0x430() [668829.177864] memory commitment underflow [668829.179049] Modules linked in: [668829.180183] CPU: 18 PID: 4414 Comm: trinity-c21 Tainted: G W 4.1.0-rc6-next-20150604-sasha-00039-g07bbbaf-dirty #2269 [668829.183509] ffff8805326b0000 0000000059024457 ffff88008d387cb8 ffffffff9fa02938 [668829.186008] 0000000000000000 ffff88008d387d38 ffff88008d387d08 ffffffff961e5336 [668829.188330] ffff8805326b1190 ffffffff965639df ffff8805326b0000 ffffed0011a70fa3 [668829.189512] Call Trace: [668829.189863] [<ffffffff9fa02938>] dump_stack+0x4f/0x7b [668829.190745] [<ffffffff961e5336>] warn_slowpath_common+0xc6/0x120 [668829.191735] [<ffffffff965639df>] ? __vm_enough_memory+0x39f/0x430 [668829.192740] [<ffffffff961e5445>] warn_slowpath_fmt+0xb5/0xf0 [668829.193588] [<ffffffff961e5390>] ? warn_slowpath_common+0x120/0x120 [668829.194554] [<ffffffff96018040>] ? arch_get_unmapped_area+0x5e0/0x5e0 [668829.195709] [<ffffffff962f08dc>] ? do_raw_spin_unlock+0x16c/0x260 [668829.196897] [<ffffffff97dae510>] ? check_preemption_disabled+0x70/0x1f0 [668829.198035] [<ffffffff965639df>] __vm_enough_memory+0x39f/0x430 [668829.199086] [<ffffffff97b052df>] security_vm_enough_memory_mm+0x7f/0xa0 [668829.200248] [<ffffffff9656ab64>] do_brk+0x3a4/0x910 [668829.201063] [<ffffffff9656b213>] ? SyS_brk+0x63/0x3e0 [668829.202033] [<ffffffff9656b213>] ? SyS_brk+0x63/0x3e0 [668829.202896] [<ffffffff9656b445>] SyS_brk+0x295/0x3e0 [668829.203701] [<ffffffff9fa6f6e2>] tracesys_phase2+0x88/0x8d [668829.205084] ---[ end trace 3d1f7fa1a382323f ]--- [668829.205871] ------------[ cut here ]------------ [668829.205883] WARNING: CPU: 21 PID: 4719 at mm/mmap.c:160 __vm_enough_memory+0x39f/0x430() [668829.205887] memory commitment underflow [668829.205890] Modules linked in: [668829.205899] CPU: 21 PID: 4719 Comm: kworker/u56:0 Tainted: G W 4.1.0-rc6-next-20150604-sasha-00039-g07bbbaf-dirty #2269 [668829.205910] ffff8808663d0000 000000009a8c3981 ffff880866207738 ffffffff9fa02938 [668829.205918] 0000000000000000 ffff8808662077b8 ffff880866207788 ffffffff961e5336 [668829.205926] ffff880866207768 ffffffff965639df ffff880a70f2455c ffffed010cc40ef3 [668829.205927] Call Trace: [668829.205938] [<ffffffff9fa02938>] dump_stack+0x4f/0x7b [668829.205948] [<ffffffff961e5336>] warn_slowpath_common+0xc6/0x120 [668829.205957] [<ffffffff965639df>] ? __vm_enough_memory+0x39f/0x430 [668829.205966] [<ffffffff961e5445>] warn_slowpath_fmt+0xb5/0xf0 [668829.205975] [<ffffffff961e5390>] ? warn_slowpath_common+0x120/0x120 [668829.205986] [<ffffffff97dae6a7>] ? debug_smp_processor_id+0x17/0x20 [668829.205996] [<ffffffff965639df>] __vm_enough_memory+0x39f/0x430 [668829.206008] [<ffffffff97b052df>] security_vm_enough_memory_mm+0x7f/0xa0 [668829.206017] [<ffffffff96568cc8>] expand_downwards+0x3f8/0xd30 [668829.206028] [<ffffffff9fa6e485>] ? _raw_spin_unlock+0x35/0x60 [668829.206039] [<ffffffff96558812>] handle_mm_fault+0x24b2/0x4440 [668829.206049] [<ffffffff962d711e>] ? trace_hardirqs_on_caller+0x2de/0x670 [668829.206057] [<ffffffff962d9920>] ? lockdep_init+0xf0/0xf0 [668829.206067] [<ffffffff96556360>] ? copy_page_range+0x1a00/0x1a00 [668829.206073] [<ffffffff962da139>] ? __lock_is_held+0xa9/0xf0 [668829.206080] [<ffffffff962da316>] ? lock_is_held+0x196/0x1f0 [668829.206091] [<ffffffff965465f7>] ? follow_page_mask+0x87/0xa70 [668829.206098] [<ffffffff965472e2>] __get_user_pages+0x302/0xfb0 [668829.206103] [<ffffffff962d9920>] ? lockdep_init+0xf0/0xf0 [668829.206112] [<ffffffff966047a0>] ? default_llseek+0x270/0x270 [668829.206122] [<ffffffff9667407a>] ? generic_getxattr+0xda/0x130 [668829.206129] [<ffffffff96546fe0>] ? follow_page_mask+0xa70/0xa70 [668829.206135] [<ffffffff962da316>] ? lock_is_held+0x196/0x1f0 [668829.206142] [<ffffffff965485b2>] get_user_pages+0x52/0x60 [668829.206151] [<ffffffff96617484>] copy_strings.isra.24+0x284/0x660 [668829.206158] [<ffffffff96617200>] ? get_user_arg_ptr.isra.20+0x70/0x70 [668829.206165] [<ffffffff966178e9>] copy_strings_kernel+0x89/0x110 [668829.206171] [<ffffffff9661ba56>] do_execveat_common.isra.28+0xfa6/0x1b10 [668829.206177] [<ffffffff9661aedd>] ? do_execveat_common.isra.28+0x42d/0x1b10 [668829.206184] [<ffffffff965c2d25>] ? arch_local_irq_restore+0x15/0x20 [668829.206197] [<ffffffff9661aab0>] ? prepare_bprm_creds+0x100/0x100 [668829.206331] [<ffffffff962da316>] ? lock_is_held+0x196/0x1f0 [668829.206342] [<ffffffff96321563>] ? rcu_read_lock_sched_held+0x1a3/0x1c0 [668829.206349] [<ffffffff965c7d08>] ? kmem_cache_alloc+0x248/0x2d0 [668829.206357] [<ffffffff965cd19a>] ? memcpy+0x3a/0x50 [668829.206363] [<ffffffff9661c5ec>] do_execve+0x2c/0x30 [668829.206372] [<ffffffff96226d1f>] ____call_usermodehelper+0x29f/0x440 [668829.206378] [<ffffffff96226a80>] ? __call_usermodehelper+0xc0/0xc0 [668829.206385] [<ffffffff9fa6fa1f>] ret_from_fork+0x3f/0x70 [668829.206392] [<ffffffff96226a80>] ? __call_usermodehelper+0xc0/0xc0
Thanks,
Sasha
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>