Skip to contentSkip to conversation

make HWLAT_DETECTOR kernel config either disabled or a module, never builtin.

PATCH In review Joakim Hernberg 3 participants 8 messages Sep 23, 2015 – Dec 22, 2015 Char and misc drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 make HWLAT_DETECTOR kernel config either disabled or a module, never builtin.

Since the HWLAT_DETECTOR driver supposedly shouldn't be builtin, but a module when used at all, it would make sense to limit the configuration choices:

---
 drivers/misc/Kconfig | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/misc/Kconfig +1 -2
index 453a616..94756c3 100644
@@ -124,8 +124,7 @@ config IBM_ASM
 
 config HWLAT_DETECTOR
 	tristate "Testing module to detect hardware-induced latencies"
-	depends on DEBUG_FS
-	depends on RING_BUFFER
+	depends on DEBUG_FS && RING_BUFFER && m
 	default m
 	---help---
 	  A simple hardware latency detector. Use this module to detect
Signature · 9 lines
2.5.3

-- 

   Joakim
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#1 lore raw
Quoted 3 lines · * Joakim Hernberg | 2015-09-23 19:33:40 [+0200]:

Since the HWLAT_DETECTOR driver supposedly shouldn't be builtin, but a module when used at all, it would make sense to limit the configuration choices:

what is wrong with it beeing built-in? The default is m already.

Quoted 2 lines

Sebastian
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
Quoted 7 lines · On Fri, 11 Dec 2015 17:52:40 +0100 Sebastian Andrzej Siewior…
Quoted 3 lines · * Joakim Hernberg | 2015-09-23 19:33:40 [+0200]:

Since the HWLAT_DETECTOR driver supposedly shouldn't be builtin, but a module when used at all, it would make sense to limit the configuration choices:

what is wrong with it beeing built-in? The default is m already.

To be quite honest I'm not quite sure anymore, think it fails when built in. I also faintly remember an irc discussion where it was claimed that it shouldn't be built into the kernel, think that was the reason for creating the patch.

Signature · 1 line
   Joakim
#3 lore raw
In reply to Sebastian Andrzej Siewior
Quoted 6 lines · On Fri, 11 Dec 2015 17:52:40 +0100 Sebastian Andrzej Siewior…
Quoted 3 lines · * Joakim Hernberg | 2015-09-23 19:33:40 [+0200]:

Since the HWLAT_DETECTOR driver supposedly shouldn't be builtin, but a module when used at all, it would make sense to limit the configuration choices:

Quoted 6 lines

what is wrong with it beeing built-in? The default is m already.

Quoted 2 lines

Sebastian

I looked at the IRC logs from when this was discussed. The hwlatdetect script looks for the kernel module and will fail if it's built-in (or compressed). Also this quote from irc: <jkacur> I need to look at the code, but offhand, I think we said, it cannot be a built-in by design

These were the reasons for the patch.

Signature · 5 lines
   Joakim
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#4 lore raw
Quoted 4 lines · * Joakim Hernberg | 2015-12-13 11:39:35 [+0100]:

I looked at the IRC logs from when this was discussed. The hwlatdetect script looks for the kernel module and will fail if it's built-in (or compressed). Also this quote from irc: <jkacur> I need to look at the code, but offhand, I think we said, it cannot be a built-in by design

Why can't it work by design in built-in mode? The module parameters enabled and threshold can be set via debugfs (and via command line). So I don't see here the difference between module and built-in. Is there something I missed?

If the hwlatdetect script is not able to handle that then it deserves to be fixed. The same goes for support of compressed modules.

Quoted 1 line

These were the reasons for the patch.

Sebastian

#5 lore raw

----- Original Message -----

Quoted 7 lines · * Joakim Hernberg | 2015-12-13 11:39:35 [+0100]:
Quoted 5 lines

I looked at the IRC logs from when this was discussed. The hwlatdetect script looks for the kernel module and will fail if it's built-in (or compressed). Also this quote from irc: <jkacur> I need to look at the code, but offhand, I think we said, it cannot be a built-in by design

Quoted 10 lines

Why can't it work by design in built-in mode? The module parameters enabled and threshold can be set via debugfs (and via command line). So I don't see here the difference between module and built-in. Is there something I missed?

If the hwlatdetect script is not able to handle that then it deserves to be fixed. The same goes for support of compressed modules.

Quoted 1 line

These were the reasons for the patch.

The script can be made to handle it, that's not a big deal. I think the historical reason was worrying that people could be accidentally running a kernel with some really heavy weight debugging that would affect performance. However, let people hang themselves if they want. Clark was going to send a patch to allow it, but probably ran out of time before the holidays. You can beat him to the punch if you want, or wait for him.

John

#6 lore raw
Quoted 4 lines · John Kacur

The script can be made to handle it, that's not a big deal. I think the historical reason was worrying that people could be accidentally running a kernel with some really heavy weight debugging that would affect performance. However, let people hang themselves if they want.

How can people hang themselves if they build it *into* the kernel? If I understand *how* this can happen I will try to address it. However from what I can see right now, nothing will happen if this module is built into the kernel.

Quoted 3 lines

John

Sebastian

#7 lore raw

----- Original Message -----

Quoted 14 lines · John Kacur
Quoted 6 lines

The script can be made to handle it, that's not a big deal. I think the historical
reason was worrying that people could be accidentally running a kernel with some really heavy weight debugging that would affect performance. However, let people
hang themselves if they want.

How can people hang themselves if they build it *into* the kernel? If I understand *how* this can happen I will try to address it. However from what I can see right now, nothing will happen if this module is built into the kernel.

Quoted 1 line
Quoted 3 lines
Quoted 2 lines

John

Sebastian

All I mean is if you are accidentally running a debug kernel, you're going to get really bad performance, which is obvious. However, why are you arguing with me, I'm agreeing with and Clark is agreeing with you, let's make it possible for people to enable it now, and forget whatever historical reason stopped us from doing that before.

John