Skip to contentSkip to conversation

get_task_comm() not exported?

PATCH In review Christoph Bartelmus 6 participants 20 messages Jan 16, 2009 – Jan 27, 2009

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

Hi,

is there any reason why get_task_comm() is not exported in fs/exec.c?

I'd like to use get_task_comm() from a kernel module and get unresolved symbol errors.

Adding EXPORT_SYMBOL(get_task_comm); to fs/exec.c fixes the problem.

Cheers,
Christoph

Please Cc.

#1 lore raw

Hi

Quoted 3 lines

Hi,

is there any reason why get_task_comm() is not exported in fs/exec.c?

In general, the only function of anybody necessarity explained is exported. if you want to export get_task_comm(), you need to explain reasonable reason.

Quoted 15 lines

I'd like to use get_task_comm() from a kernel module and get unresolved symbol errors.

Adding EXPORT_SYMBOL(get_task_comm); to fs/exec.c fixes the problem.

Cheers,
Christoph

Please Cc.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#2 lore raw

Hi,

Quoted 5 lines · KOSAKI Motohiro
Quoted 1 line

is there any reason why get_task_comm() is not exported in fs/exec.c?

In general, the only function of anybody necessarity explained is exported. if you want to export get_task_comm(), you need to explain reasonable reason.

It's nothing that important: just want to print the executable name to the logs during error handling. get_task_comm() is the required accessor function.

http://lirc.cvs.sourceforge.net/viewvc/lirc/lirc/drivers/lirc_dev/lirc_dev.c?revision=1.70&view=markup

Cheers,
Christoph

#3 lore raw

PATCH 1/1 export get_task_comm()

Quoted 9 lines

Hi,

Quoted 5 lines · KOSAKI Motohiro
Quoted 1 line

is there any reason why get_task_comm() is not exported in fs/exec.c?

In general, the only function of anybody necessarity explained is exported. if you want to export get_task_comm(), you need to explain reasonable reason.

Quoted 5 lines

It's nothing that important: just want to print the executable name to the logs during error handling. get_task_comm() is the required accessor function.

http://lirc.cvs.sourceforge.net/viewvc/lirc/lirc/drivers/lirc_dev/lirc_dev.c?revision=1.70&view=markup

To be honest, I don't use get_task_comm(). but I don't oppose your request.

==
Subject: [PATCH] export get_task_comm()

task::comm is good debugging information and driver developer want to use this information easily.

---
 fs/exec.c |    1 +
 1 file changed, 1 insertion(+)
fs/exec.c +1 -0
Index: b/fs/exec.c
===================================================================
@@ -936,6 +936,7 @@ char *get_task_comm(char *buf, struct ta
 	task_unlock(tsk);
 	return buf;
 }
+EXPORT_SYMBOL_GPL(get_task_comm);
 
 void set_task_comm(struct task_struct *tsk, char *buf)
 {
#4 lore raw

Re: [PATCH] export get_task_comm()

Quoted 33 lines · On Tue, 27 Jan 2009 14:09:05 +0900 (JST) KOSAKI Motohiro…
Quoted 9 lines

Hi,

Quoted 5 lines · KOSAKI Motohiro
Quoted 1 line

is there any reason why get_task_comm() is not exported in fs/exec.c?

In general, the only function of anybody necessarity explained is exported. if you want to export get_task_comm(), you need to explain reasonable reason.

Quoted 5 lines

It's nothing that important: just want to print the executable name to the logs during error handling. get_task_comm() is the required accessor function.

http://lirc.cvs.sourceforge.net/viewvc/lirc/lirc/drivers/lirc_dev/lirc_dev.c?revision=1.70&view=markup

To be honest, I don't use get_task_comm(). but I don't oppose your request.

==
Subject: [PATCH] export get_task_comm()

task::comm is good debugging information and driver developer want to use this information easily.

---
 fs/exec.c |    1 +
 1 file changed, 1 insertion(+)

Index: b/fs/exec.c
===================================================================

fs/exec.c char *get_task_comm(char *buf, struct ta · line 942

Quoted 7 lines
fs/exec.c +1 -0
@@ -936,6 +936,7 @@ char *get_task_comm(char *buf, struct ta
 	task_unlock(tsk);
 	return buf;
 }
+EXPORT_SYMBOL_GPL(get_task_comm);
 
 void set_task_comm(struct task_struct *tsk, char *buf)
 {

Ho hum, I suppose so. I redid the changelog a bit:

task_struct.comm[] is useful for debugging and driver developers want to use this information easily. Direct access to task_struct.comm[] is a bit racy, so export the official accessor.

#5 lore raw

Re: [PATCH] export get_task_comm()

Quoted 6 lines · Andrew Morton

Ho hum, I suppose so. I redid the changelog a bit:

task_struct.comm[] is useful for debugging and driver developers want to use this information easily. Direct access to task_struct.comm[] is a bit racy, so export the official accessor.

Maybe lirc should be submitted to staging/ before we go exporting symbols for out of tree things... ;-)

#6 lore raw

Re: [PATCH] export get_task_comm()

Quoted 10 lines · On Tue, 27 Jan 2009 00:19:19 -0500 Kyle McMartin <[email protected]> wrote:
Quoted 6 lines · Andrew Morton

Ho hum, I suppose so. I redid the changelog a bit:

task_struct.comm[] is useful for debugging and driver developers want to use this information easily. Direct access to task_struct.comm[] is a bit racy, so export the official accessor.

Maybe lirc should be submitted to staging/ before we go exporting symbols for out of tree things... ;-)

y:/usr/src/linux-2.6.29-rc2> grep -r 'current->comm' drivers | wc -l 77

:(

#7 lore raw

Re: [PATCH] export get_task_comm()

Quoted 10 lines · On Mon, 26 Jan 2009 21:26:15 -0800 Andrew Morton <[email protected]>…
Quoted 8 lines · On Tue, 27 Jan 2009 00:19:19 -0500 Kyle McMartin <[email protected]> wrote:
Quoted 6 lines · Andrew Morton

Ho hum, I suppose so. I redid the changelog a bit:

task_struct.comm[] is useful for debugging and driver developers want to use this information easily. Direct access to task_struct.comm[] is a bit racy, so export the official accessor.

Quoted 7 lines
Quoted 2 lines

Maybe lirc should be submitted to staging/ before we go exporting symbols for out of tree things... ;-)

y:/usr/src/linux-2.6.29-rc2> grep -r 'current->comm' drivers | wc -l 77

:(

It's worth a checkpatch rule, I guess: "direct access to task_struct.comm is racy - use get_task_comm()".

#8 lore raw

PATCH 1/1 make checkpatch warn about access to current->comm

From: Kyle McMartin <[email protected]>

Suggest using the get_task_comm accessor versus direct access to current->comm.

---

I just mashed it in the middle of existing checks to minimize risk of collisions.

scripts/checkpatch.pl +5 -0
index 45eb0ae..4f2da5d 100755
@@ -2533,6 +2533,11 @@ sub process {
 				$herecurr);
 		}
 
+# direct access to current task name is racy, suggest accessor instead.
+		if ($line =~ /current\-\>comm/) {
+			WARN("direct access to current->comm is racy, use get_task_comm() instead.\n" . $herecurr);
+		}
+
 # use of NR_CPUS is usually wrong
 # ignore definitions of NR_CPUS and usage to define arrays as likely right
 		if ($line =~ /\bNR_CPUS\b/ &&
#9 lore raw

Re: [PATCH] make checkpatch warn about access to current->comm

scripts/checkpatch.pl sub process { · line 2543

Quoted 8 lines
scripts/checkpatch.pl +5 -0
+# direct access to current task name is racy, suggest accessor instead.
+		if ($line =~ /current\-\>comm/) {
+			WARN("direct access to current->comm is racy, use get_task_comm() instead.\n" . $herecurr);
+		}
+
 # use of NR_CPUS is usually wrong
 # ignore definitions of NR_CPUS and usage to define arrays as likely right
 		if ($line =~ /\bNR_CPUS\b/ &&

./scripts/checkpatch --file fs/exec.c
------------------------------------------------------
WARNING: direct access to current->comm is racy, use get_task_comm() instead. #952: FILE: exec.c:952:
+ char tcomm[sizeof(current->comm)];

(snip)

WARNING: direct access to current->comm is racy, use get_task_comm() instead. #1459: FILE: exec.c:1459:
+ "%s", current->comm);

(snip)

WARNING: direct access to current->comm is racy, use get_task_comm() instead. #1788: FILE: exec.c:1788:
+ if (!strcmp(delimit, current->comm)) {

---------------------------

I think "char tcomm[sizeof(current->comm)];" is valid code. if checkpatch.pl don't warn "sizeof(current->comm)", I'm glad.

#10 lore raw

Re: [PATCH] make checkpatch warn about access to current->comm

Quoted 3 lines · KOSAKI Motohiro

I think "char tcomm[sizeof(current->comm)];" is valid code. if checkpatch.pl don't warn "sizeof(current->comm)", I'm glad.

Awfully interesting way of writing TASK_COMM_LEN :)

cheers, Kyle

#11 lore raw

Re: [PATCH] make checkpatch warn about access to current->comm

Quoted 5 lines · KOSAKI Motohiro
Quoted 2 lines

I think "char tcomm[sizeof(current->comm)];" is valid code. if checkpatch.pl don't warn "sizeof(current->comm)", I'm glad.

Awfully interesting way of writing TASK_COMM_LEN :)

I don't think so awfully.
I think "sizeof(array_val)" is typical code in kernel.

I agree that we can rewrite s/tcomm[sizeof(current->comm)]/char tcomm[TASK_COMM_LEN]/. but it's annoyed.

#12 lore raw

Re: [PATCH] make checkpatch warn about access to current->comm

Quoted 4 lines · KOSAKI Motohiro
Quoted 4 lines · KOSAKI Motohiro
Quoted 2 lines

I think "char tcomm[sizeof(current->comm)];" is valid code. if checkpatch.pl don't warn "sizeof(current->comm)", I'm glad.

Quoted 8 lines
Quoted 1 line

Awfully interesting way of writing TASK_COMM_LEN :)

I don't think so awfully.
I think "sizeof(array_val)" is typical code in kernel.

I agree that we can rewrite s/tcomm[sizeof(current->comm)]/char tcomm[TASK_COMM_LEN]/. but it's annoyed.

If current->comm was changed to be kmalloc'd this would need to be changed anyway, so why not just do it now?

regards, Kyle

#16 lore raw

Re: [PATCH] export get_task_comm()

In reply to Andrew Morton
Quoted 15 lines · Andrew Morton
Quoted 13 lines · On Mon, 26 Jan 2009 21:26:15 -0800 Andrew Morton <[email protected]>…
Quoted 10 lines · On Tue, 27 Jan 2009 00:19:19 -0500 Kyle McMartin <[email protected]> wrote:
Quoted 6 lines · Andrew Morton

Ho hum, I suppose so. I redid the changelog a bit:

task_struct.comm[] is useful for debugging and driver developers want to use this information easily. Direct access to task_struct.comm[] is a bit racy, so export the official accessor.

Maybe lirc should be submitted to staging/ before we go exporting symbols for out of tree things... ;-)

Quoted 7 lines
Quoted 4 lines

y:/usr/src/linux-2.6.29-rc2> grep -r 'current->comm' drivers | wc -l 77

:(

It's worth a checkpatch rule, I guess: "direct access to task_struct.comm is racy - use get_task_comm()".

And majority of usages is some debugging printk where nobody cares if ->comm corrupted.

Changelog says is useful for debugging. That's right, tsk->comm is useful for debugging, not allocating temporary buffer + get_task_comm().

Some ->comm usages are for kernel threads which never change it, for starters.

current->comm is always safe, because, current is not executing prctl(2)!

I'd say nothing should be done and, heavens forbid, adding this to checkpatch.pl.

#17 lore raw

Re: [PATCH] export get_task_comm()

Quoted 27 lines · On Tue, 27 Jan 2009 11:49:56 +0300 Alexey Dobriyan <[email protected]> wrote:
Quoted 15 lines · Andrew Morton
Quoted 13 lines · On Mon, 26 Jan 2009 21:26:15 -0800 Andrew Morton <[email protected]>…
Quoted 10 lines · On Tue, 27 Jan 2009 00:19:19 -0500 Kyle McMartin <[email protected]> wrote:
Quoted 6 lines · Andrew Morton

Ho hum, I suppose so. I redid the changelog a bit:

task_struct.comm[] is useful for debugging and driver developers want to use this information easily. Direct access to task_struct.comm[] is a bit racy, so export the official accessor.

Maybe lirc should be submitted to staging/ before we go exporting symbols for out of tree things... ;-)

Quoted 7 lines
Quoted 4 lines

y:/usr/src/linux-2.6.29-rc2> grep -r 'current->comm' drivers | wc -l 77

:(

It's worth a checkpatch rule, I guess: "direct access to task_struct.comm is racy - use get_task_comm()".

And majority of usages is some debugging printk where nobody cares if ->comm corrupted.

Quoted 6 lines

Changelog says is useful for debugging. That's right, tsk->comm is useful for debugging, not allocating temporary buffer + get_task_comm().

Some ->comm usages are for kernel threads which never change it, for starters.

current->comm is always safe, because, current is not executing prctl(2)!

That's a good point.

Quoted 1 line

I'd say nothing should be done and, heavens forbid, adding this to checkpatch.pl.

hm, yeah, OK, I'll drop it.

#18 lore raw

Re: [PATCH] export get_task_comm()

Quoted 11 lines
Quoted 11 lines
Quoted 2 lines

It's worth a checkpatch rule, I guess: "direct access to task_struct.comm is racy - use get_task_comm()".

And majority of usages is some debugging printk where nobody cares if ->comm corrupted.

Changelog says is useful for debugging. That's right, tsk->comm is useful for debugging, not allocating temporary buffer + get_task_comm().

Some ->comm usages are for kernel threads which never change it, for starters.

Quoted 7 lines
Quoted 1 line

current->comm is always safe, because, current is not executing prctl(2)!

That's a good point.

Quoted 1 line

I'd say nothing should be done and, heavens forbid, adding this to checkpatch.pl.

hm, yeah, OK, I'll drop it.

Oh, I have to agree Alexey because he is always right ;)

#19 lore raw

Re: [PATCH] export get_task_comm()

In reply to Andrew Morton
Quoted 9 lines · Andrew Morton
Quoted 3 lines

task::comm is good debugging information and driver developer want to use this information easily.

Ho hum, I suppose so. I redid the changelog a bit:

task_struct.comm[] is useful for debugging and driver developers want to use this information easily. Direct access to task_struct.comm[] is a bit racy, so export the official accessor.

The biggest issue I have with this is that the whole "get_task_comm()" interface is not very good for random users - it inherently depends on the result buffer being at least sizeof(tsk->comm).

If we export it to random routines, I get the feeling that we should pass in the size of the result buffer, so that they don't have to know about this requirement.

Linus

#13 lore raw

Re: [PATCH] make checkpatch warn about access to current->comm

In reply to Kyle McMartin
Quoted 3 lines · Kyle McMartin

Suggest using the get_task_comm accessor versus direct access to current->comm.

I think "current->comm" is fine, and not racy.

It only gets racy when you ask for the name of _another_ task.

And quite frankly, I don't think anybody but /proc does that anyway. I think this whole "get_task_comm()" thing is overrated. Most people are better off doing just "current->comm".

Linus

#14 lore raw

Re: [PATCH] make checkpatch warn about access to current->comm

Quoted 7 lines · Linus Torvalds
Quoted 3 lines · Kyle McMartin

Suggest using the get_task_comm accessor versus direct access to current->comm.

Quoted 8 lines

I think "current->comm" is fine, and not racy.

It only gets racy when you ask for the name of _another_ task.

And quite frankly, I don't think anybody but /proc does that anyway. I think this whole "get_task_comm()" thing is overrated. Most people are better off doing just "current->comm".

Sure, fine by me. I'd forgotten that prctl doesn't have a `pid' argument to change another tasks comm.

regards, Kyle

#15 lore raw

Re: [PATCH] make checkpatch warn about access to current->comm

In reply to Linus Torvalds
Quoted 5 lines · Linus Torvalds
Quoted 3 lines · Kyle McMartin

Suggest using the get_task_comm accessor versus direct access to current->comm.

Quoted 7 lines

I think "current->comm" is fine, and not racy.

It only gets racy when you ask for the name of _another_ task.

And quite frankly, I don't think anybody but /proc does that anyway. I think this whole "get_task_comm()" thing is overrated. Most people are better off doing just "current->comm".

This issue only came up because for someone like me it's not obvious at all that using "current->comm" is safe and the comment in sched.h explicitly points out that task_struct.comm should be accessed with [gs]et_task_comm.

Christoph