'struct pid' is a "variable sized struct" - a header with an array
of upids at the end.
A size of the array depends on a level (depth) of pid namespaces. Now
a level of pidns is not limited, so 'struct pid' can be more than one
page.
Looks reasonable, that it should be less than a page. MAX_PIS_NS_LEVEL
is not calculated from PAGE_SIZE, because in this case it depends on
architectures, config options and it will be reduced, if someone adds a
new fields in struct pid or struct upid.
I suggest to set MAX_PIS_NS_LEVEL = 32, because it saves ability to
expand "struct pid" and it's more than enough for all known for me
use-cases. When someone finds a reasonable use case, we can add a
config option or a sysctl parameter.
In addition it will reduce effect of another problem, when we have many
nested namespaces and the oldest one starts dying. zap_pid_ns_processe
will be called for each namespace and find_vpid will be called for each
process in a namespace. find_vpid will be called minimum max_level^2 / 2
times. The reason of that is that when we found a bit in pidmap, we
can't determine this pidns is top for this process or it isn't.
vpid is a heavy operation, so a fork bomb, which create many nested
namespace, can do a system inaccessible for a long time.
Cc: Andrew Morton <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Cyrill Gorcunov <[email protected]>
Cc: "Eric W. Biederman" <[email protected]>
Cc: Pavel Emelyanov <[email protected]>
Signed-off-by: Andrew Vagin <[email protected]>
---
kernel/pid_namespace.c | 6 ++++++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index b051fa6..598bfb3 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -70,12 +70,18 @@ err_alloc:
return NULL;
}
+/* MAX_PID_NS_LEVEL is needed for limiting size of 'struct pid' */
+#define MAX_PID_NS_LEVEL 32
+
static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_pid_ns)
{
struct pid_namespace *ns;
unsigned int level = parent_pid_ns->level + 1;
int i, err = -ENOMEM;
+ if (level > MAX_PID_NS_LEVEL)
+ goto out;
+
ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
if (ns == NULL)
goto out;
--
1.7.1
Hello Andrew and Oleg,
Andrew, what do you think about this patch? I reworked it according
with your comments to the previous version.
Oleg, could you send Ack in this version, if it's ok for you.
Thanks.
2012/10/12 Andrew Vagin <[email protected]>:
> 'struct pid' is a "variable sized struct" - a header with an array
> of upids at the end.
>
> A size of the array depends on a level (depth) of pid namespaces. Now
> a level of pidns is not limited, so 'struct pid' can be more than one
> page.
>
> Looks reasonable, that it should be less than a page. MAX_PIS_NS_LEVEL
> is not calculated from PAGE_SIZE, because in this case it depends on
> architectures, config options and it will be reduced, if someone adds a
> new fields in struct pid or struct upid.
>
> I suggest to set MAX_PIS_NS_LEVEL = 32, because it saves ability to
> expand "struct pid" and it's more than enough for all known for me
> use-cases. When someone finds a reasonable use case, we can add a
> config option or a sysctl parameter.
>
> In addition it will reduce effect of another problem, when we have many
> nested namespaces and the oldest one starts dying. zap_pid_ns_processe
> will be called for each namespace and find_vpid will be called for each
> process in a namespace. find_vpid will be called minimum max_level^2 / 2
> times. The reason of that is that when we found a bit in pidmap, we
> can't determine this pidns is top for this process or it isn't.
>
> vpid is a heavy operation, so a fork bomb, which create many nested
> namespace, can do a system inaccessible for a long time.
>
> Cc: Andrew Morton <[email protected]>
> Cc: Oleg Nesterov <[email protected]>
> Cc: Cyrill Gorcunov <[email protected]>
> Cc: "Eric W. Biederman" <[email protected]>
> Cc: Pavel Emelyanov <[email protected]>
> Signed-off-by: Andrew Vagin <[email protected]>
> ---
> kernel/pid_namespace.c | 6 ++++++
> 1 files changed, 6 insertions(+), 0 deletions(-)
>
> diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
> index b051fa6..598bfb3 100644
> --- a/kernel/pid_namespace.c
> +++ b/kernel/pid_namespace.c
> @@ -70,12 +70,18 @@ err_alloc:
> return NULL;
> }
>
> +/* MAX_PID_NS_LEVEL is needed for limiting size of 'struct pid' */
> +#define MAX_PID_NS_LEVEL 32
> +
> static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_pid_ns)
> {
> struct pid_namespace *ns;
> unsigned int level = parent_pid_ns->level + 1;
> int i, err = -ENOMEM;
>
> + if (level > MAX_PID_NS_LEVEL)
> + goto out;
> +
> ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
> if (ns == NULL)
> goto out;
> --
> 1.7.1
>
On 10/12, Andrew Vagin wrote:
>
> 'struct pid' is a "variable sized struct" - a header with an array
> of upids at the end.
>
> A size of the array depends on a level (depth) of pid namespaces. Now
> a level of pidns is not limited, so 'struct pid' can be more than one
> page.
>
> Looks reasonable, that it should be less than a page. MAX_PIS_NS_LEVEL
> is not calculated from PAGE_SIZE, because in this case it depends on
> architectures, config options and it will be reduced, if someone adds a
> new fields in struct pid or struct upid.
>
> I suggest to set MAX_PIS_NS_LEVEL = 32, because it saves ability to
> expand "struct pid" and it's more than enough for all known for me
> use-cases. When someone finds a reasonable use case, we can add a
> config option or a sysctl parameter.
>
> In addition it will reduce effect of another problem, when we have many
> nested namespaces and the oldest one starts dying. zap_pid_ns_processe
> will be called for each namespace and find_vpid will be called for each
> process in a namespace. find_vpid will be called minimum max_level^2 / 2
> times. The reason of that is that when we found a bit in pidmap, we
> can't determine this pidns is top for this process or it isn't.
>
> vpid is a heavy operation, so a fork bomb, which create many nested
> namespace, can do a system inaccessible for a long time.
>
> Cc: Andrew Morton <[email protected]>
> Cc: Oleg Nesterov <[email protected]>
> Cc: Cyrill Gorcunov <[email protected]>
> Cc: "Eric W. Biederman" <[email protected]>
> Cc: Pavel Emelyanov <[email protected]>
> Signed-off-by: Andrew Vagin <[email protected]>
Acked-by: Oleg Nesterov <[email protected]>
> ---
> kernel/pid_namespace.c | 6 ++++++
> 1 files changed, 6 insertions(+), 0 deletions(-)
>
> diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
> index b051fa6..598bfb3 100644
> --- a/kernel/pid_namespace.c
> +++ b/kernel/pid_namespace.c
> @@ -70,12 +70,18 @@ err_alloc:
> return NULL;
> }
>
> +/* MAX_PID_NS_LEVEL is needed for limiting size of 'struct pid' */
> +#define MAX_PID_NS_LEVEL 32
> +
> static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_pid_ns)
> {
> struct pid_namespace *ns;
> unsigned int level = parent_pid_ns->level + 1;
> int i, err = -ENOMEM;
>
> + if (level > MAX_PID_NS_LEVEL)
> + goto out;
> +
> ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
> if (ns == NULL)
> goto out;
> --
> 1.7.1
>
On Fri, 12 Oct 2012 16:30:42 +0400
Andrew Vagin <[email protected]> wrote:
> 'struct pid' is a "variable sized struct" - a header with an array
> of upids at the end.
>
> A size of the array depends on a level (depth) of pid namespaces. Now
> a level of pidns is not limited, so 'struct pid' can be more than one
> page.
>
> Looks reasonable, that it should be less than a page. MAX_PIS_NS_LEVEL
> is not calculated from PAGE_SIZE, because in this case it depends on
> architectures, config options and it will be reduced, if someone adds a
> new fields in struct pid or struct upid.
>
> I suggest to set MAX_PIS_NS_LEVEL = 32, because it saves ability to
> expand "struct pid" and it's more than enough for all known for me
> use-cases. When someone finds a reasonable use case, we can add a
> config option or a sysctl parameter.
>
> In addition it will reduce effect of another problem, when we have many
> nested namespaces and the oldest one starts dying. zap_pid_ns_processe
> will be called for each namespace and find_vpid will be called for each
> process in a namespace. find_vpid will be called minimum max_level^2 / 2
> times. The reason of that is that when we found a bit in pidmap, we
> can't determine this pidns is top for this process or it isn't.
>
> vpid is a heavy operation, so a fork bomb, which create many nested
> namespace, can do a system inaccessible for a long time.
>
> --- a/kernel/pid_namespace.c
> +++ b/kernel/pid_namespace.c
> @@ -70,12 +70,18 @@ err_alloc:
> return NULL;
> }
>
> +/* MAX_PID_NS_LEVEL is needed for limiting size of 'struct pid' */
> +#define MAX_PID_NS_LEVEL 32
> +
> static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_pid_ns)
> {
> struct pid_namespace *ns;
> unsigned int level = parent_pid_ns->level + 1;
> int i, err = -ENOMEM;
>
> + if (level > MAX_PID_NS_LEVEL)
> + goto out;
> +
> ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
> if (ns == NULL)
> goto out;
hm, OK. This will kind-of fix the free_pid_ns()-excessive-recursion
issue which we already fixed by other means ;)
I don't think this problem is serious enough to bother backporting into
-stable but I guess we can/should do it in 3.7. Agree?
I think that returning -ENOMEM in response to an excessive nesting
attempt is misleading - the system *didn't* run out of memory. EINVAL
is better?
From: Andrew Morton <[email protected]>
Subject: pidns-limit-the-nesting-depth-of-pid-namespaces-fix
return -EINVAL in response to excessive nesting, not -ENOMEM
Cc: "Eric W. Biederman" <[email protected]>
Cc: Andrew Vagin <[email protected]>
Cc: Cyrill Gorcunov <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Pavel Emelyanov <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
---
kernel/pid_namespace.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff -puN kernel/pid_namespace.c~pidns-limit-the-nesting-depth-of-pid-namespaces-fix kernel/pid_namespace.c
--- a/kernel/pid_namespace.c~pidns-limit-the-nesting-depth-of-pid-namespaces-fix
+++ a/kernel/pid_namespace.c
@@ -78,11 +78,15 @@ static struct pid_namespace *create_pid_
{
struct pid_namespace *ns;
unsigned int level = parent_pid_ns->level + 1;
- int i, err = -ENOMEM;
+ int i;
+ int err;
- if (level > MAX_PID_NS_LEVEL)
+ if (level > MAX_PID_NS_LEVEL) {
+ err = -EINVAL;
goto out;
+ }
+ err = -ENOMEM;
ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
if (ns == NULL)
goto out;
_
2012/10/24 Andrew Morton <[email protected]>:
> On Fri, 12 Oct 2012 16:30:42 +0400
> Andrew Vagin <[email protected]> wrote:
>
>> 'struct pid' is a "variable sized struct" - a header with an array
>> of upids at the end.
>>
>> A size of the array depends on a level (depth) of pid namespaces. Now
>> a level of pidns is not limited, so 'struct pid' can be more than one
>> page.
>>
>> Looks reasonable, that it should be less than a page. MAX_PIS_NS_LEVEL
>> is not calculated from PAGE_SIZE, because in this case it depends on
>> architectures, config options and it will be reduced, if someone adds a
>> new fields in struct pid or struct upid.
>>
>> I suggest to set MAX_PIS_NS_LEVEL = 32, because it saves ability to
>> expand "struct pid" and it's more than enough for all known for me
>> use-cases. When someone finds a reasonable use case, we can add a
>> config option or a sysctl parameter.
>>
>> In addition it will reduce effect of another problem, when we have many
>> nested namespaces and the oldest one starts dying. zap_pid_ns_processe
>> will be called for each namespace and find_vpid will be called for each
>> process in a namespace. find_vpid will be called minimum max_level^2 / 2
>> times. The reason of that is that when we found a bit in pidmap, we
>> can't determine this pidns is top for this process or it isn't.
>>
>> vpid is a heavy operation, so a fork bomb, which create many nested
>> namespace, can do a system inaccessible for a long time.
>>
>> --- a/kernel/pid_namespace.c
>> +++ b/kernel/pid_namespace.c
>> @@ -70,12 +70,18 @@ err_alloc:
>> return NULL;
>> }
>>
>> +/* MAX_PID_NS_LEVEL is needed for limiting size of 'struct pid' */
>> +#define MAX_PID_NS_LEVEL 32
>> +
>> static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_pid_ns)
>> {
>> struct pid_namespace *ns;
>> unsigned int level = parent_pid_ns->level + 1;
>> int i, err = -ENOMEM;
>>
>> + if (level > MAX_PID_NS_LEVEL)
>> + goto out;
>> +
>> ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
>> if (ns == NULL)
>> goto out;
>
> hm, OK. This will kind-of fix the free_pid_ns()-excessive-recursion
> issue which we already fixed by other means ;)
I caught both problems with help of an one program, which create many
nested namespaces. Actually this patch prevents another problem. A few
thousand nested namespaces is destroyed for more than one minutes and
in this period a system is inaccessible.
>
> I don't think this problem is serious enough to bother backporting into
> -stable but I guess we can/should do it in 3.7. Agree?
Yes.
>
> I think that returning -ENOMEM in response to an excessive nesting
> attempt is misleading - the system *didn't* run out of memory. EINVAL
> is better?
I chose ENOMEM by analogy with max_pid. When a new PID can not be
allocated, ENOMEM is returned too.
>
>
>
> From: Andrew Morton <[email protected]>
> Subject: pidns-limit-the-nesting-depth-of-pid-namespaces-fix
>
> return -EINVAL in response to excessive nesting, not -ENOMEM
>
> Cc: "Eric W. Biederman" <[email protected]>
> Cc: Andrew Vagin <[email protected]>
> Cc: Cyrill Gorcunov <[email protected]>
> Cc: Oleg Nesterov <[email protected]>
> Cc: Pavel Emelyanov <[email protected]>
> Signed-off-by: Andrew Morton <[email protected]>
> ---
>
> kernel/pid_namespace.c | 8 ++++++--
> 1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff -puN kernel/pid_namespace.c~pidns-limit-the-nesting-depth-of-pid-namespaces-fix kernel/pid_namespace.c
> --- a/kernel/pid_namespace.c~pidns-limit-the-nesting-depth-of-pid-namespaces-fix
> +++ a/kernel/pid_namespace.c
> @@ -78,11 +78,15 @@ static struct pid_namespace *create_pid_
> {
> struct pid_namespace *ns;
> unsigned int level = parent_pid_ns->level + 1;
> - int i, err = -ENOMEM;
> + int i;
> + int err;
>
> - if (level > MAX_PID_NS_LEVEL)
> + if (level > MAX_PID_NS_LEVEL) {
> + err = -EINVAL;
> goto out;
> + }
>
> + err = -ENOMEM;
> ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
> if (ns == NULL)
> goto out;
> _
>
On Wed, 24 Oct 2012 09:38:57 +0400
Andrey Wagin <[email protected]> wrote:
> >
> > I think that returning -ENOMEM in response to an excessive nesting
> > attempt is misleading - the system *didn't* run out of memory. EINVAL
> > is better?
>
> I chose ENOMEM by analogy with max_pid. When a new PID can not be
> allocated, ENOMEM is returned too.
I don't know what this means - please be carefully specific when
identifying kernel code.
If you're referring to kernel/pid.c:alloc_pid() then -ENOMEM is
appropriate there, because a failure *is* caused by memory allocation
failure.
But ENOMEM isn't appropriate for nesting-depth-exceeded - we shouldn't
tell the user "you ran out of memory" when he didn't! -EINVAL isn't
really appropriate either ("Invalid argument") but it has become a
general you-screwed-up catchall and seems to me to be the most
appropriate errno we have available.
2012/10/24 Andrew Morton <[email protected]>:
> On Wed, 24 Oct 2012 09:38:57 +0400
> Andrey Wagin <[email protected]> wrote:
>
>> >
>> > I think that returning -ENOMEM in response to an excessive nesting
>> > attempt is misleading - the system *didn't* run out of memory. EINVAL
>> > is better?
>>
>> I chose ENOMEM by analogy with max_pid. When a new PID can not be
>> allocated, ENOMEM is returned too.
>
> I don't know what this means - please be carefully specific when
> identifying kernel code.
Sorry.
>
> If you're referring to kernel/pid.c:alloc_pid() then -ENOMEM is
> appropriate there, because a failure *is* caused by memory allocation
> failure.
I'm referring to alloc_pidmap().
For example I set pid_max to 500 and try to create more than 500 processes.
[pid 345] clone(child_stack=0,
flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD,
child_tidptr=0x7f8721716a10) = -1 ENOMEM (Cannot allocate memory)
Actually I'm agree with EINVAL and a patch is attached to this message.
Thanks.
>
> But ENOMEM isn't appropriate for nesting-depth-exceeded - we shouldn't
> tell the user "you ran out of memory" when he didn't! -EINVAL isn't
> really appropriate either ("Invalid argument") but it has become a
> general you-screwed-up catchall and seems to me to be the most
> appropriate errno we have available.
>