Skip to contentSkip to conversation

mdb: export task_curr

PATCH In review [email protected] 4 participants 11 messages Aug 14, 2008 – Aug 15, 2008

lore
Share & tools
lore
#0 lore raw

PATCH 26/29 mdb: export task_curr

export the task_curr function to the module based kernel debugger to enable process back tracing and state display.

kernel/sched.c +3 -0
@@ -1736,6 +1736,9 @@
 {
 	return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif
 
 static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
 {
#1 lore raw
Quoted 8 lines · On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]> wrote:

export the task_curr function to the module based kernel debugger to enable process back tracing and state display.

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 6 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core kernel code.

#2 lore raw
Quoted 7 lines · On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]> wrote:
Quoted 6 lines

export the task_curr function to the module based kernel debugger to enable
process back tracing and state display.

Quoted 13 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 9 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core kernel code.

Well,then please suggest how a kernel debugger can be module based and still be able to get this information some other way that's generic and minimal impact.

Jeff

#3 lore raw
Quoted 21 lines · [email protected] writes:
Quoted 7 lines · On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]> wrote:
Quoted 6 lines

export the task_curr function to the module based kernel debugger to enable
process back tracing and state display.

Quoted 13 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 9 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core kernel code.

Quoted 3 lines

Well,then please suggest how a kernel debugger can be module based and still be able to get this information some other way that's generic and minimal impact.

EXPORT_SYMBOL_GPL(task_curr);

Without the #if's. Just export it.

Hannes

#4 lore raw
Quoted 31 lines
Quoted 23 lines · [email protected] writes:
Quoted 8 lines
Quoted 6 lines · On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]> wrote:

export the task_curr function to the module based kernel debugger to enable
process back tracing and state display.

Quoted 14 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 9 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core kernel
code.

Quoted 5 lines

Well,then please suggest how a kernel debugger can be module based and still
be able to get this information some other way that's generic and minimal
impact.

Quoted 6 lines

EXPORT_SYMBOL_GPL(task_curr);

Without the #if's. Just export it.

Hannes

OK. That's simple and minimal.

Jeff

#5 lore raw
In reply to [email protected]
Quoted 21 lines · * [email protected] <[email protected]> wrote:
Quoted 7 lines · On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]> wrote:
Quoted 6 lines

export the task_curr function to the module based kernel debugger to enable
process back tracing and state display.

Quoted 13 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 9 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core kernel code.

Quoted 3 lines

Well,then please suggest how a kernel debugger can be module based and still be able to get this information some other way that's generic and minimal impact.

FYI, there's a built-in kernel debugger in the upstream kernel already:
kernel/kgdb.c - and it does not need that export.

So if you are interested in kernel debuggers i'd suggest to work with the KGDB folks to extend it with whatever feature-set is missing.

They are friendly, very easy to work with and are open to the thousands-of-years-old scientific method of not duplicating effort, working together, going forward gradually, etc.

Ingo

#6 lore raw
Quoted 33 lines · * [email protected] <[email protected]> wrote:
Quoted 20 lines
Quoted 2 lines

On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]>

Quoted 6 lines · wrote:
Quoted 6 lines

export the task_curr function to the module based kernel debugger to enable
process back tracing and state display.

Quoted 11 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 9 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core

Quoted 7 lines

kernel

Quoted 2 lines

code.

Well,then please suggest how a kernel debugger can be module based and still be able to get this information some other way that's generic and minimal impact.

FYI, there's a built-in kernel debugger in the upstream kernel already:
kernel/kgdb.c - and it does not need that export.

Quoted 9 lines

So if you are interested in kernel debuggers i'd suggest to work with the KGDB folks to extend it with whatever feature-set is missing.

They are friendly, very easy to work with and are open to the thousands-of-years-old scientific method of not duplicating effort, working together, going forward gradually, etc.

Ingo

Sorry, but I have my own path and direction, and its not as a tin can on a
strong tied to the tail of kdb or kgdb. Again, it matters not what's in
Linus' tree or not, but whether this is useful to others and they use it.

Quoted 1 line

From what I have seen based on the non-stop downloads from my FTP server,

its needed for certain classes of users -- me included. I will continue to enhance it, release it, and move it forward.

kdb and kgdb can go their way. I'm certain my code is being reviewed as we speak and if its useful to them, so much the better.

Jeff

#7 lore raw
Quoted 36 lines · * [email protected] <[email protected]> wrote:
Quoted 36 lines · * [email protected] <[email protected]> wrote:
Quoted 20 lines
Quoted 2 lines

On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]>

Quoted 6 lines · wrote:
Quoted 6 lines

export the task_curr function to the module based kernel debugger to enable
process back tracing and state display.

Quoted 11 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 9 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core

Quoted 7 lines

kernel

Quoted 2 lines

code.

Well,then please suggest how a kernel debugger can be module based and still be able to get this information some other way that's generic and minimal impact.

FYI, there's a built-in kernel debugger in the upstream kernel already:
kernel/kgdb.c - and it does not need that export.

So if you are interested in kernel debuggers i'd suggest to work with the KGDB folks to extend it with whatever feature-set is missing.

Quoted 8 lines
Quoted 6 lines

They are friendly, very easy to work with and are open to the thousands-of-years-old scientific method of not duplicating effort, working together, going forward gradually, etc.

Ingo

Sorry, but I have my own path and direction, [...]

doing things alone is your unalienable personal right, and since thus you apparently didnt intend your patch to be included in the Linux kernel that's as far as my interest in this as a maintainer goes.

Thanks,

Ingo

#8 lore raw
Quoted 52 lines · * [email protected] <[email protected]> wrote:
Quoted 41 lines
Quoted 7 lines · * [email protected] <[email protected]> wrote:
Quoted 4 lines
Quoted 2 lines

On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]>

Quoted 1 line · wrote:
Quoted 1 line

export the task_curr function to the module based kernel debugger

to

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

enable
process back tracing and state display.

Signed-off-by Jeffrey Vernon Merkey

([email protected])

Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 10 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core

Quoted 5 lines
Quoted 5 lines

kernel

Quoted 2 lines

code.

Well,then please suggest how a kernel debugger can be module based

and

Quoted 4 lines
Quoted 2 lines

still be able to get this information some other way that's generic and minimal impact.

FYI, there's a built-in kernel debugger in the upstream kernel

already:

Quoted 5 lines

kernel/kgdb.c - and it does not need that export.

So if you are interested in kernel debuggers i'd suggest to work with the KGDB folks to extend it with whatever feature-set is missing.

Quoted 8 lines
Quoted 6 lines

They are friendly, very easy to work with and are open to the thousands-of-years-old scientific method of not duplicating effort, working together, going forward gradually, etc.

Ingo

Sorry, but I have my own path and direction, [...]

Quoted 8 lines

doing things alone is your unalienable personal right, and since thus you apparently didnt intend your patch to be included in the Linux kernel that's as far as my interest in this as a maintainer goes.

Thanks,

Ingo

Ingo,

The patches are submitted so you can stop putting words in my mouth. If it goes in, it goes in -- its submitted. Your comment about this export and comparing it to kgdb demonstrates 1) you have never even looked at the code of either debugger in this area 2) you were unaware I did not need this exported except as a module 3) which kgdb and kdb do not support either.

A maintainer would be expected to actually review the code. Its clear from the comments above you have not, so I see little interest anyway from you, so who cares.

Jeff

#9 lore raw
Quoted 53 lines
Quoted 53 lines · * [email protected] <[email protected]> wrote:
Quoted 50 lines
Quoted 3 lines

* [email protected] <[email protected]>

Quoted 4 lines · wrote:
Quoted 4 lines
Quoted 2 lines

On Thu, Aug 14, 2008 at 9:14 AM, <[email protected]>

Quoted 1 line · wrote:
Quoted 1 line

export the task_curr function to the module based kernel debugger

to

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

enable
process back tracing and state display.

Signed-off-by Jeffrey Vernon Merkey

([email protected])

Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 line 1741

Quoted 10 lines
kernel/sched.c 2008-08-13 11:56:03.000000000 -0600 renamed from kernel/sched.c 2008-08-13 14:22:32.000000000 -0600 +3 -0
@@ -1736,6 +1736,9 @@
 {
       return cpu_curr(task_cpu(p)) == p;
 }
+#if defined(CONFIG_MDB_MODULE)
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core

Quoted 5 lines
Quoted 5 lines

kernel

Quoted 2 lines

code.

Well,then please suggest how a kernel debugger can be module based

and

Quoted 4 lines
Quoted 2 lines

still be able to get this information some other way that's generic and minimal impact.

FYI, there's a built-in kernel debugger in the upstream kernel

already:

Quoted 11 lines

kernel/kgdb.c - and it does not need that export.

So if you are interested in kernel debuggers i'd suggest to work with the KGDB folks to extend it with whatever feature-set is missing.

They are friendly, very easy to work with and are open to the thousands-of-years-old scientific method of not duplicating effort, working together, going forward gradually, etc.

Ingo

Sorry, but I have my own path and direction, [...]

Quoted 8 lines
Quoted 8 lines

doing things alone is your unalienable personal right, and since thus you apparently didnt intend your patch to be included in the Linux kernel that's as far as my interest in this as a maintainer goes.

Thanks,

Ingo

I removed the export from this file and will resubmit the patch series. This will remove you from the decision list for this patch series. What other files do you claim to maintain so I can make certain remove them from my patches?

I can get the variable some other way.

:-)

Thanks

Jeff

#10 lore raw
In reply to Ingo Molnar
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

kernel/sched.c line 1741

Quoted 3 lines
kernel/sched.c +2 -0
+EXPORT_SYMBOL_GPL(task_curr);
+#endif

We usually don't export symbols conditionally, especially in core

Quoted 7 lines
Quoted 4 lines

kernel

Quoted 2 lines

code.

FYI, there's a built-in kernel debugger in the upstream kernel already:
kernel/kgdb.c - and it does not need that export.

I do not need this export either -- unless mdb runs as a kernel module - which kdb and kgdb cannot do today ...

Jeff

Quoted 3 lines

Ingo