Re: [PATCH v2 1/1] mm/memory-failure: Poison read receives SIGKILL instead of SIGBUS if mmaped more than once
On 7/24/2019 4:43 PM, Naoya Horiguchi wrote:
On Wed, Jul 24, 2019 at 04:33:23PM -0600, Jane Chu wrote:
Mmap /dev/dax more than once, then read the poison location using address
from one of the mappings. The other mappings due to not having the page
mapped in will cause SIGKILLs delivered to the process. SIGKILL succeeds
over SIGBUS, so user process looses the opportunity to handle the UE.
Although one may add MAP_POPULATE to mmap(2) to work around the issue,
MAP_POPULATE makes mapping 128GB of pmem several magnitudes slower, so
isn't always an option.
Details -
ndctl inject-error --block=10 --count=1 namespace6.0
./read_poison -x dax6.0 -o 5120 -m 2
mmaped address 0x7f5bb660
mmaped address 0x7f3cf360
doing local read at address 0x7f3cf3601400
Killed
Console messages in instrumented kernel -
mce: Uncorrected hardware memory error in user-access at edbe201400
Memory failure: tk->addr = 7f5bb6601000
Memory failure: address edbe201: call dev_pagemap_mapping_shift
dev_pagemap_mapping_shift: page edbe201: no PUD
Memory failure: tk->size_shift == 0
Memory failure: Unable to find user space address edbe201 in read_poison
Memory failure: tk->addr = 7f3cf3601000
Memory failure: address edbe201: call dev_pagemap_mapping_shift
Memory failure: tk->size_shift = 21
Memory failure: 0xedbe201: forcibly killing read_poison:22434 because of
failure to unmap corrupted page
=> to deliver SIGKILL
Memory failure: 0xedbe201: Killing read_poison:22434 due to hardware memory
corruption
=> to deliver SIGBUS
Signed-off-by: Jane Chu
Suggested-by: Naoya Horiguchi
---
mm/memory-failure.c | 62 ++---
1 file changed, 26 insertions(+), 36 deletions(-)
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index d9cc660..bd4db33 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -199,7 +199,6 @@ struct to_kill {
struct task_struct *tsk;
unsigned long addr;
short size_shift;
- char addr_valid;
};
/*
@@ -304,43 +303,43 @@ static unsigned long dev_pagemap_mapping_shift(struct
page *page,
/*
* Schedule a process for later kill.
* Uses GFP_ATOMIC allocations to avoid potential recursions in the VM.
- * TBD would GFP_NOIO be enough?
*/
static void add_to_kill(struct task_struct *tsk, struct page *p,
struct vm_area_struct *vma,
- struct list_head *to_kill,
- struct to_kill **tkc)
+ struct list_head *to_kill)
{
struct to_kill *tk;
- if (*tkc) {
- tk = *tkc;
- *tkc = NULL;
- } else {
- tk = kmalloc(sizeof(struct to_kill), GFP_ATOMIC);
- if (!tk) {
- pr_err("Memory failure: Out of memory while machine check
handling\n");
- return;
- }
+ tk = kmalloc(sizeof(struct to_kill), GFP_ATOMIC);
+ if (!tk) {
+ pr_err("Memory failure: Out of memory while machine check
handling\n");
+ return;
As Dan pointed out, the cleanup part can be delivered as a separate patch.
My bad, will take care splitting up the patch.
}
+
tk->addr = page_address_in_vma(p, vma);
- tk->addr_valid = 1;
if (is_zone_device_page(p))
tk->size_shift = dev_pagemap_mapping_shift(p, vma);
else
tk->size_shift = compound_order(compound_head(p)) + PAGE_SHIFT;
/*
-* In theory we don't have to kill when the page was
-* munmaped. But it could be also a mremap. Since that's
-* likely very rare kill anyways just out of paranoia, but use
-* a SIGKILL because the error is not contained anymore.
+* Send SIGKILL if "tk->addr == -EFAULT". Also, as
+* "tk->size_shift" is always non-zero for !is_zone_device_page(),
+* so "tk->size_shift == 0" effectively checks no mapping on
+* ZONE_DEVICE. Indeed, when a devdax page is mmapped N times
+* to a process' address space, it's possible not all N VMAs
+* contain mappings for the page, but at least one VMA does.
+* Only deliver SIGBUS with payload derived from the VMA that
+* has a mapping for the page.
OK, so SIGBUSs are sent M times (where M is the number of mappings
for the page). Then I'm convinced that we need "else if" block below.
Yes. I run read_poison that mmaps /dev/dax 4 times with MAPS_POPULATE flag
set, so the kernel attempted sending SIGBUS 4 times.
One time, while the poison was consumed at uaddr[1] (2nd mmap), but the
SIGBUS payload indicated the si_addr was uaddr[3] (4th mmap).
thanks!
-jane
Thanks,
Naoya Horiguchi
*/
- if (tk->addr == -EFAULT || tk->size_shift == 0) {
+ if (tk->addr == -EFAULT) {
pr_info("Memory failure: Unable to find user space address %lx in
%s\n",
page_to_pfn(p), tsk->comm);
-
Re: [PATCH v2 1/1] mm/memory-failure: Poison read receives SIGKILL instead of SIGBUS if mmaped more than once
On Wed, Jul 24, 2019 at 04:33:23PM -0600, Jane Chu wrote:
> Mmap /dev/dax more than once, then read the poison location using address
> from one of the mappings. The other mappings due to not having the page
> mapped in will cause SIGKILLs delivered to the process. SIGKILL succeeds
> over SIGBUS, so user process looses the opportunity to handle the UE.
>
> Although one may add MAP_POPULATE to mmap(2) to work around the issue,
> MAP_POPULATE makes mapping 128GB of pmem several magnitudes slower, so
> isn't always an option.
>
> Details -
>
> ndctl inject-error --block=10 --count=1 namespace6.0
>
> ./read_poison -x dax6.0 -o 5120 -m 2
> mmaped address 0x7f5bb660
> mmaped address 0x7f3cf360
> doing local read at address 0x7f3cf3601400
> Killed
>
> Console messages in instrumented kernel -
>
> mce: Uncorrected hardware memory error in user-access at edbe201400
> Memory failure: tk->addr = 7f5bb6601000
> Memory failure: address edbe201: call dev_pagemap_mapping_shift
> dev_pagemap_mapping_shift: page edbe201: no PUD
> Memory failure: tk->size_shift == 0
> Memory failure: Unable to find user space address edbe201 in read_poison
> Memory failure: tk->addr = 7f3cf3601000
> Memory failure: address edbe201: call dev_pagemap_mapping_shift
> Memory failure: tk->size_shift = 21
> Memory failure: 0xedbe201: forcibly killing read_poison:22434 because of
> failure to unmap corrupted page
> => to deliver SIGKILL
> Memory failure: 0xedbe201: Killing read_poison:22434 due to hardware memory
> corruption
> => to deliver SIGBUS
>
> Signed-off-by: Jane Chu
> Suggested-by: Naoya Horiguchi
> ---
> mm/memory-failure.c | 62
> ++---
> 1 file changed, 26 insertions(+), 36 deletions(-)
>
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index d9cc660..bd4db33 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -199,7 +199,6 @@ struct to_kill {
> struct task_struct *tsk;
> unsigned long addr;
> short size_shift;
> - char addr_valid;
> };
>
> /*
> @@ -304,43 +303,43 @@ static unsigned long dev_pagemap_mapping_shift(struct
> page *page,
> /*
> * Schedule a process for later kill.
> * Uses GFP_ATOMIC allocations to avoid potential recursions in the VM.
> - * TBD would GFP_NOIO be enough?
> */
> static void add_to_kill(struct task_struct *tsk, struct page *p,
> struct vm_area_struct *vma,
> -struct list_head *to_kill,
> -struct to_kill **tkc)
> +struct list_head *to_kill)
> {
> struct to_kill *tk;
>
> - if (*tkc) {
> - tk = *tkc;
> - *tkc = NULL;
> - } else {
> - tk = kmalloc(sizeof(struct to_kill), GFP_ATOMIC);
> - if (!tk) {
> - pr_err("Memory failure: Out of memory while machine
> check handling\n");
> - return;
> - }
> + tk = kmalloc(sizeof(struct to_kill), GFP_ATOMIC);
> + if (!tk) {
> + pr_err("Memory failure: Out of memory while machine check
> handling\n");
> + return;
As Dan pointed out, the cleanup part can be delivered as a separate patch.
> }
> +
> tk->addr = page_address_in_vma(p, vma);
> - tk->addr_valid = 1;
> if (is_zone_device_page(p))
> tk->size_shift = dev_pagemap_mapping_shift(p, vma);
> else
> tk->size_shift = compound_order(compound_head(p)) + PAGE_SHIFT;
>
> /*
> - * In theory we don't have to kill when the page was
> - * munmaped. But it could be also a mremap. Since that's
> - * likely very rare kill anyways just out of paranoia, but use
> - * a SIGKILL because the error is not contained anymore.
> + * Send SIGKILL if "tk->addr == -EFAULT". Also, as
> + * "tk->size_shift" is always non-zero for !is_zone_device_page(),
> + * so "tk->size_shift == 0" effectively checks no mapping on
> + * ZONE_DEVICE. Indeed, when a devdax page is mmapped N times
> + * to a process' address space, it's possible not all N VMAs
> + * contain mappings for the page, but at least one VMA does.
> + * Only deliver SIGBUS with payload derived from the VMA that
> + * has a mapping for the page.
OK, so SIGBUSs are sent M times (where M is the number of mappings
for the page). Then I'm convinced that we need "else if" block below.
Thanks,
Naoya Horiguchi
>*/
> - if (tk->addr == -EFAULT || tk->size_shift == 0) {
> + if (tk->addr == -EFAULT) {
> pr_info("Memory failure: Unable to find user space address %lx
> in %s\n",
> page_to_pfn(p), tsk->comm);
> - tk->addr_valid = 0;
> + } else if (tk->size_shift == 0) {
> + kfree(tk);
> + return;
> }
> +
> get_task_struct(tsk);
> tk->tsk = tsk;
> list_add_tail(>nd, to_kill);
> @@ -366,7 +365,7 @@ static
[PATCH v2 1/1] mm/memory-failure: Poison read receives SIGKILL instead of SIGBUS if mmaped more than once
Mmap /dev/dax more than once, then read the poison location using address
from one of the mappings. The other mappings due to not having the page
mapped in will cause SIGKILLs delivered to the process. SIGKILL succeeds
over SIGBUS, so user process looses the opportunity to handle the UE.
Although one may add MAP_POPULATE to mmap(2) to work around the issue,
MAP_POPULATE makes mapping 128GB of pmem several magnitudes slower, so
isn't always an option.
Details -
ndctl inject-error --block=10 --count=1 namespace6.0
./read_poison -x dax6.0 -o 5120 -m 2
mmaped address 0x7f5bb660
mmaped address 0x7f3cf360
doing local read at address 0x7f3cf3601400
Killed
Console messages in instrumented kernel -
mce: Uncorrected hardware memory error in user-access at edbe201400
Memory failure: tk->addr = 7f5bb6601000
Memory failure: address edbe201: call dev_pagemap_mapping_shift
dev_pagemap_mapping_shift: page edbe201: no PUD
Memory failure: tk->size_shift == 0
Memory failure: Unable to find user space address edbe201 in read_poison
Memory failure: tk->addr = 7f3cf3601000
Memory failure: address edbe201: call dev_pagemap_mapping_shift
Memory failure: tk->size_shift = 21
Memory failure: 0xedbe201: forcibly killing read_poison:22434 because of
failure to unmap corrupted page
=> to deliver SIGKILL
Memory failure: 0xedbe201: Killing read_poison:22434 due to hardware memory
corruption
=> to deliver SIGBUS
Signed-off-by: Jane Chu
Suggested-by: Naoya Horiguchi
---
mm/memory-failure.c | 62 ++---
1 file changed, 26 insertions(+), 36 deletions(-)
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index d9cc660..bd4db33 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -199,7 +199,6 @@ struct to_kill {
struct task_struct *tsk;
unsigned long addr;
short size_shift;
- char addr_valid;
};
/*
@@ -304,43 +303,43 @@ static unsigned long dev_pagemap_mapping_shift(struct
page *page,
/*
* Schedule a process for later kill.
* Uses GFP_ATOMIC allocations to avoid potential recursions in the VM.
- * TBD would GFP_NOIO be enough?
*/
static void add_to_kill(struct task_struct *tsk, struct page *p,
struct vm_area_struct *vma,
- struct list_head *to_kill,
- struct to_kill **tkc)
+ struct list_head *to_kill)
{
struct to_kill *tk;
- if (*tkc) {
- tk = *tkc;
- *tkc = NULL;
- } else {
- tk = kmalloc(sizeof(struct to_kill), GFP_ATOMIC);
- if (!tk) {
- pr_err("Memory failure: Out of memory while machine
check handling\n");
- return;
- }
+ tk = kmalloc(sizeof(struct to_kill), GFP_ATOMIC);
+ if (!tk) {
+ pr_err("Memory failure: Out of memory while machine check
handling\n");
+ return;
}
+
tk->addr = page_address_in_vma(p, vma);
- tk->addr_valid = 1;
if (is_zone_device_page(p))
tk->size_shift = dev_pagemap_mapping_shift(p, vma);
else
tk->size_shift = compound_order(compound_head(p)) + PAGE_SHIFT;
/*
-* In theory we don't have to kill when the page was
-* munmaped. But it could be also a mremap. Since that's
-* likely very rare kill anyways just out of paranoia, but use
-* a SIGKILL because the error is not contained anymore.
+* Send SIGKILL if "tk->addr == -EFAULT". Also, as
+* "tk->size_shift" is always non-zero for !is_zone_device_page(),
+* so "tk->size_shift == 0" effectively checks no mapping on
+* ZONE_DEVICE. Indeed, when a devdax page is mmapped N times
+* to a process' address space, it's possible not all N VMAs
+* contain mappings for the page, but at least one VMA does.
+* Only deliver SIGBUS with payload derived from the VMA that
+* has a mapping for the page.
*/
- if (tk->addr == -EFAULT || tk->size_shift == 0) {
+ if (tk->addr == -EFAULT) {
pr_info("Memory failure: Unable to find user space address %lx
in %s\n",
page_to_pfn(p), tsk->comm);
- tk->addr_valid = 0;
+ } else if (tk->size_shift == 0) {
+ kfree(tk);
+ return;
}
+
get_task_struct(tsk);
tk->tsk = tsk;
list_add_tail(>nd, to_kill);
@@ -366,7 +365,7 @@ static void kill_procs(struct list_head *to_kill, int
forcekill, bool fail,
* make sure the process doesn't catch the
* signal and then access the memory. Just kill it.
*/
- if (fail || tk->addr_valid == 0) {
+ if (fail || tk->addr == -EFAULT) {
pr_err("Memory failure: %#lx: forcibly killing
