LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH] rlim in proc/<pid>/status
@ 2008-01-15 10:06 Clifford Wolf
  2008-01-15 10:47 ` KOSAKI Motohiro
  2008-01-15 20:36 ` [PATCH] " serge
  0 siblings, 2 replies; 7+ messages in thread
From: Clifford Wolf @ 2008-01-15 10:06 UTC (permalink / raw)
  To: lkml

Hi,

because I needed it already twice in two different projects this week: the
following patch adds rlim (ulimits) output to /proc/<pid>/status.

Please let me know if there is another (already existing) way of accessing
this information easy (i.e. connecting with gdb to the process in question
and 'injecting' a getrlimit() call does not count.. ;-).

yours,
 - clifford

Signed-off-by: Clifford Wolf <clifford@clifford.at>

--- linux/fs/proc/array.c	(revision 757)
+++ linux/fs/proc/array.c	(working copy)
@@ -239,6 +239,55 @@
 	}
 }
 
+static char *rlim_names[RLIM_NLIMITS] = {
+	[RLIMIT_CPU]        = "CPU",
+	[RLIMIT_FSIZE]      = "FSize",
+	[RLIMIT_DATA]       = "Data",
+	[RLIMIT_STACK]      = "Stack",
+	[RLIMIT_CORE]       = "Core",
+	[RLIMIT_RSS]        = "RSS",
+	[RLIMIT_NPROC]      = "NProc",
+	[RLIMIT_NOFILE]     = "NoFile",
+	[RLIMIT_MEMLOCK]    = "MemLock",
+	[RLIMIT_AS]         = "AddrSpace",
+	[RLIMIT_LOCKS]      = "Locks",
+	[RLIMIT_SIGPENDING] = "SigPending",
+	[RLIMIT_MSGQUEUE]   = "MsgQueue",
+	[RLIMIT_NICE]       = "Nice",
+	[RLIMIT_RTPRIO]     = "RTPrio"
+};
+
+static inline char *task_rlim(struct task_struct *p, char *buffer)
+{
+	unsigned long flags;
+	struct rlimit rlim[RLIM_NLIMITS];
+	int i;
+	
+	rcu_read_lock();
+	if (lock_task_sighand(p, &flags)) {
+		for (i=0; i<RLIM_NLIMITS; i++)
+			rlim[i] = p->signal->rlim[i];
+	}
+	rcu_read_unlock();
+
+	for (i=0; i<RLIM_NLIMITS; i++) {
+		if (rlim_names[i])
+			buffer += sprintf(buffer, "Rlim%s:\t", rlim_names[i]);
+		else
+			buffer += sprintf(buffer, "Rlim%d:\t", i);
+		if (rlim[i].rlim_cur != ~0)
+			buffer += sprintf(buffer, "%lu\t", rlim[i].rlim_cur);
+		else
+			buffer += sprintf(buffer, "-\t");
+		if (rlim[i].rlim_max != ~0)
+			buffer += sprintf(buffer, "%lu\n", rlim[i].rlim_max);
+		else
+			buffer += sprintf(buffer, "-\n");
+	}
+
+	return buffer;
+}
+
 static inline char *task_sig(struct task_struct *p, char *buffer)
 {
 	unsigned long flags;
@@ -310,6 +359,7 @@
 		buffer = task_mem(mm, buffer);
 		mmput(mm);
 	}
+	buffer = task_rlim(task, buffer);
 	buffer = task_sig(task, buffer);
 	buffer = task_cap(task, buffer);
 	buffer = cpuset_task_status_allowed(task, buffer);

-- 
[..] If it still doesn't work, re-write it in assembler. This won't fix the
bug, but it will make sure no one else finds it and makes you look bad.

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: rlim in proc/<pid>/status
  2008-01-15 10:06 [PATCH] rlim in proc/<pid>/status Clifford Wolf
@ 2008-01-15 10:47 ` KOSAKI Motohiro
  2008-01-15 12:15   ` Clifford Wolf
  2008-01-15 20:36 ` [PATCH] " serge
  1 sibling, 1 reply; 7+ messages in thread
From: KOSAKI Motohiro @ 2008-01-15 10:47 UTC (permalink / raw)
  To: Clifford Wolf; +Cc: kosaki.motohiro, lkml

Hi

sound good for me.
a few question please.

> +	for (i=0; i<RLIM_NLIMITS; i++) {
> +		if (rlim_names[i])
> +			buffer += sprintf(buffer, "Rlim%s:\t", rlim_names[i]);
> +		else
> +			buffer += sprintf(buffer, "Rlim%d:\t", i);

this else is really necessary?

> +		if (rlim[i].rlim_cur != ~0)
> +			buffer += sprintf(buffer, "%lu\t", rlim[i].rlim_cur);
> +		else
> +			buffer += sprintf(buffer, "-\t");
> +		if (rlim[i].rlim_max != ~0)
> +			buffer += sprintf(buffer, "%lu\n", rlim[i].rlim_max);
> +		else
> +			buffer += sprintf(buffer, "-\n");

Why do you don't use RLIM_INFINITY?

- kosaki



^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: rlim in proc/<pid>/status
  2008-01-15 10:47 ` KOSAKI Motohiro
@ 2008-01-15 12:15   ` Clifford Wolf
  0 siblings, 0 replies; 7+ messages in thread
From: Clifford Wolf @ 2008-01-15 12:15 UTC (permalink / raw)
  To: KOSAKI Motohiro; +Cc: lkml

Hi,

On Tue, Jan 15, 2008 at 07:47:22PM +0900, KOSAKI Motohiro wrote:
> sound good for me.
> a few question please.
> 
> > +	for (i=0; i<RLIM_NLIMITS; i++) {
> > +		if (rlim_names[i])
> > +			buffer += sprintf(buffer, "Rlim%s:\t", rlim_names[i]);
> > +		else
> > +			buffer += sprintf(buffer, "Rlim%d:\t", i);
> 
> this else is really necessary?

no. not with the current sources. maybe something like the following would
be better:

#if RLIM_NLIMITS != 15
#  error New RLIM_NLIMITS add mising entries to rlim_names[]
#endif

> > +		if (rlim[i].rlim_cur != ~0)
> > +			buffer += sprintf(buffer, "%lu\t", rlim[i].rlim_cur);
> > +		else
> > +			buffer += sprintf(buffer, "-\t");
> > +		if (rlim[i].rlim_max != ~0)
> > +			buffer += sprintf(buffer, "%lu\n", rlim[i].rlim_max);
> > +		else
> > +			buffer += sprintf(buffer, "-\n");
> 
> Why do you don't use RLIM_INFINITY?

because I'm blind and didn't see it... ;-)

maybe it would also be better to output 'inf' instead of '-' in this case?

yours,
 - clifford

-- 
The number of the beast - vi vi vi.

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH] rlim in proc/<pid>/status
  2008-01-15 10:06 [PATCH] rlim in proc/<pid>/status Clifford Wolf
  2008-01-15 10:47 ` KOSAKI Motohiro
@ 2008-01-15 20:36 ` serge
  2008-01-16  7:03   ` [PATCH] rlim in proc/<pid>/status (2nd rev.) Clifford Wolf
  1 sibling, 1 reply; 7+ messages in thread
From: serge @ 2008-01-15 20:36 UTC (permalink / raw)
  To: Clifford Wolf; +Cc: lkml

Quoting Clifford Wolf (clifford@clifford.at):
> Hi,
> 
> because I needed it already twice in two different projects this week: the
> following patch adds rlim (ulimits) output to /proc/<pid>/status.
> 
> Please let me know if there is another (already existing) way of accessing
> this information easy (i.e. connecting with gdb to the process in question
> and 'injecting' a getrlimit() call does not count.. ;-).
> 
> yours,
>  - clifford
> 
> Signed-off-by: Clifford Wolf <clifford@clifford.at>
> 
> --- linux/fs/proc/array.c	(revision 757)
> +++ linux/fs/proc/array.c	(working copy)
> @@ -239,6 +239,55 @@
>  	}
>  }
>  
> +static char *rlim_names[RLIM_NLIMITS] = {
> +	[RLIMIT_CPU]        = "CPU",
> +	[RLIMIT_FSIZE]      = "FSize",
> +	[RLIMIT_DATA]       = "Data",
> +	[RLIMIT_STACK]      = "Stack",
> +	[RLIMIT_CORE]       = "Core",
> +	[RLIMIT_RSS]        = "RSS",
> +	[RLIMIT_NPROC]      = "NProc",
> +	[RLIMIT_NOFILE]     = "NoFile",
> +	[RLIMIT_MEMLOCK]    = "MemLock",
> +	[RLIMIT_AS]         = "AddrSpace",
> +	[RLIMIT_LOCKS]      = "Locks",
> +	[RLIMIT_SIGPENDING] = "SigPending",
> +	[RLIMIT_MSGQUEUE]   = "MsgQueue",
> +	[RLIMIT_NICE]       = "Nice",
> +	[RLIMIT_RTPRIO]     = "RTPrio"
> +};
> +
> +static inline char *task_rlim(struct task_struct *p, char *buffer)
> +{
> +	unsigned long flags;
> +	struct rlimit rlim[RLIM_NLIMITS];
> +	int i;
> +	
> +	rcu_read_lock();
> +	if (lock_task_sighand(p, &flags)) {
> +		for (i=0; i<RLIM_NLIMITS; i++)
> +			rlim[i] = p->signal->rlim[i];

I'm confused - where do you unlock_task_sighand()?

> +	}
> +	rcu_read_unlock();
> +
> +	for (i=0; i<RLIM_NLIMITS; i++) {
> +		if (rlim_names[i])
> +			buffer += sprintf(buffer, "Rlim%s:\t", rlim_names[i]);
> +		else
> +			buffer += sprintf(buffer, "Rlim%d:\t", i);
> +		if (rlim[i].rlim_cur != ~0)
> +			buffer += sprintf(buffer, "%lu\t", rlim[i].rlim_cur);
> +		else
> +			buffer += sprintf(buffer, "-\t");
> +		if (rlim[i].rlim_max != ~0)
> +			buffer += sprintf(buffer, "%lu\n", rlim[i].rlim_max);
> +		else
> +			buffer += sprintf(buffer, "-\n");
> +	}
> +
> +	return buffer;
> +}
> +
>  static inline char *task_sig(struct task_struct *p, char *buffer)
>  {
>  	unsigned long flags;
> @@ -310,6 +359,7 @@
>  		buffer = task_mem(mm, buffer);
>  		mmput(mm);
>  	}
> +	buffer = task_rlim(task, buffer);
>  	buffer = task_sig(task, buffer);
>  	buffer = task_cap(task, buffer);
>  	buffer = cpuset_task_status_allowed(task, buffer);
> 
> -- 
> [..] If it still doesn't work, re-write it in assembler. This won't fix the
> bug, but it will make sure no one else finds it and makes you look bad.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH] rlim in proc/<pid>/status (2nd rev.)
  2008-01-15 20:36 ` [PATCH] " serge
@ 2008-01-16  7:03   ` Clifford Wolf
  2008-01-16  7:33     ` KOSAKI Motohiro
  0 siblings, 1 reply; 7+ messages in thread
From: Clifford Wolf @ 2008-01-16  7:03 UTC (permalink / raw)
  To: serge; +Cc: lkml

Hi,

On Tue, Jan 15, 2008 at 02:36:59PM -0600, serge@hallyn.com wrote:
> > +	rcu_read_lock();
> > +	if (lock_task_sighand(p, &flags)) {
> > +		for (i=0; i<RLIM_NLIMITS; i++)
> > +			rlim[i] = p->signal->rlim[i];
> 
> I'm confused - where do you unlock_task_sighand()?

oh fsck! thanks for that pointer..

Here is a new version of the patch which solves this issue and the issues
adressed earlier in this thread by kosaki.

yours,
 - clifford

Signed-off-by: Clifford Wolf <clifford@clifford.at>

--- linux/fs/proc/array.c	(revision 750)
+++ linux/fs/proc/array.c	(revision 764)
@@ -239,6 +239,58 @@
 	}
 }
 
+static char *rlim_names[RLIM_NLIMITS] = {
+	[RLIMIT_CPU]        = "CPU",
+	[RLIMIT_FSIZE]      = "FSize",
+	[RLIMIT_DATA]       = "Data",
+	[RLIMIT_STACK]      = "Stack",
+	[RLIMIT_CORE]       = "Core",
+	[RLIMIT_RSS]        = "RSS",
+	[RLIMIT_NPROC]      = "NProc",
+	[RLIMIT_NOFILE]     = "NoFile",
+	[RLIMIT_MEMLOCK]    = "MemLock",
+	[RLIMIT_AS]         = "AddrSpace",
+	[RLIMIT_LOCKS]      = "Locks",
+	[RLIMIT_SIGPENDING] = "SigPending",
+	[RLIMIT_MSGQUEUE]   = "MsgQueue",
+	[RLIMIT_NICE]       = "Nice",
+	[RLIMIT_RTPRIO]     = "RTPrio"
+};
+
+#if RLIM_NLIMITS != 15
+#  error Value of RLIM_NLIMITS changed. \
+         Please update rlim_names in fs/proc/array.c
+#endif
+
+static inline char *task_rlim(struct task_struct *p, char *buffer)
+{
+	unsigned long flags;
+	struct rlimit rlim[RLIM_NLIMITS];
+	int i;
+	
+	rcu_read_lock();
+	if (lock_task_sighand(p, &flags)) {
+		for (i=0; i<RLIM_NLIMITS; i++)
+			rlim[i] = p->signal->rlim[i];
+		unlock_task_sighand(p, &flags);
+	}
+	rcu_read_unlock();
+
+	for (i=0; i<RLIM_NLIMITS; i++) {
+		buffer += sprintf(buffer, "Rlim%s:\t", rlim_names[i]);
+		if (rlim[i].rlim_cur != RLIM_INFINITY)
+			buffer += sprintf(buffer, "%lu\t", rlim[i].rlim_cur);
+		else
+			buffer += sprintf(buffer, "inf\t");
+		if (rlim[i].rlim_max != RLIM_INFINITY)
+			buffer += sprintf(buffer, "%lu\n", rlim[i].rlim_max);
+		else
+			buffer += sprintf(buffer, "inf\n");
+	}
+
+	return buffer;
+}
+
 static inline char *task_sig(struct task_struct *p, char *buffer)
 {
 	unsigned long flags;
@@ -310,6 +362,7 @@
 		buffer = task_mem(mm, buffer);
 		mmput(mm);
 	}
+	buffer = task_rlim(task, buffer);
 	buffer = task_sig(task, buffer);
 	buffer = task_cap(task, buffer);
 	buffer = cpuset_task_status_allowed(task, buffer);

-- 
perl -le '$_=1;(1x$_)!~/^(11+)\1+$/&&print${_}while$_++<1000'|fmt

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH] rlim in proc/<pid>/status (2nd rev.)
  2008-01-16  7:03   ` [PATCH] rlim in proc/<pid>/status (2nd rev.) Clifford Wolf
@ 2008-01-16  7:33     ` KOSAKI Motohiro
  2008-01-16 10:04       ` Clifford Wolf
  0 siblings, 1 reply; 7+ messages in thread
From: KOSAKI Motohiro @ 2008-01-16  7:33 UTC (permalink / raw)
  To: Clifford Wolf; +Cc: kosaki.motohiro, serge, lkml

Hi Clifford,

> +static inline char *task_rlim(struct task_struct *p, char *buffer)
> +{
> +	unsigned long flags;
> +	struct rlimit rlim[RLIM_NLIMITS];
> +	int i;
> +	
> +	rcu_read_lock();
> +	if (lock_task_sighand(p, &flags)) {
> +		for (i=0; i<RLIM_NLIMITS; i++)
> +			rlim[i] = p->signal->rlim[i];
> +		unlock_task_sighand(p, &flags);
> +	}

lock_task_sighand is possible return NULL?
if so, rlim is uninitialized when NULL.


- kosaki



^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH] rlim in proc/<pid>/status (2nd rev.)
  2008-01-16  7:33     ` KOSAKI Motohiro
@ 2008-01-16 10:04       ` Clifford Wolf
  0 siblings, 0 replies; 7+ messages in thread
From: Clifford Wolf @ 2008-01-16 10:04 UTC (permalink / raw)
  To: KOSAKI Motohiro; +Cc: lkml

Hi,

On Wed, Jan 16, 2008 at 04:33:12PM +0900, KOSAKI Motohiro wrote:
> Hi Clifford,
> 
> > +static inline char *task_rlim(struct task_struct *p, char *buffer)
> > +{
> > +	unsigned long flags;
> > +	struct rlimit rlim[RLIM_NLIMITS];
> > +	int i;
> > +	
> > +	rcu_read_lock();
> > +	if (lock_task_sighand(p, &flags)) {
> > +		for (i=0; i<RLIM_NLIMITS; i++)
> > +			rlim[i] = p->signal->rlim[i];
> > +		unlock_task_sighand(p, &flags);
> > +	}
> 
> lock_task_sighand is possible return NULL?
> if so, rlim is uninitialized when NULL.

hmm.. can p->sighand be NULL here? If so, there also is a problem in the
task_sig() function in the same file.

In fact that code is copy&paste from task_sig(). In fact I'm not even sure
if I actually need to lock anything to access p->signal->rlim.. I've
searched the kernel code a bit and it looks like most acesses to the rlimits
are done unlocked, which is no problem imo given the typical change-pattern
of this values..

anyone a clue on this issues?

yours,
 - clifford

-- 
2B OR (NOT 2B) That is the question. The answer is FF.

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2008-01-16 10:10 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-01-15 10:06 [PATCH] rlim in proc/<pid>/status Clifford Wolf
2008-01-15 10:47 ` KOSAKI Motohiro
2008-01-15 12:15   ` Clifford Wolf
2008-01-15 20:36 ` [PATCH] " serge
2008-01-16  7:03   ` [PATCH] rlim in proc/<pid>/status (2nd rev.) Clifford Wolf
2008-01-16  7:33     ` KOSAKI Motohiro
2008-01-16 10:04       ` Clifford Wolf

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).