Skip to contentSkip to conversation

2.6.2: "-" or "_", thats the question

DISCUSSION Ryan Reich 4 participants 5 messages Feb 16, 2004 – Feb 17, 2004 8 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

Ryan Reich

author
#8 lore raw
Quoted 10 lines · Harald Dunkel
Quoted 8 lines · Ryan Reich

On Mon, 16 Feb 2004, Ryan Reich wrote:

Sorry, I didn't realize that your problem was also the inconsistency in module names. Someone else suggested using a shell expansion; you could try

cat /proc/modules | tr _ - | grep -q "^${module_name/_/-}"

Quoted 7 lines
Quoted 2 lines

which is both short and works.

tr is usually in /usr/bin, which might not be available at boot time. And probably you mean 'grep -q "^${module_name//_/-}"'

Did I mention that the inconsistency requires ugly and error- prone workarounds? QED

It's not THAT ugly, and it's also error-prone to type at the keyboard. The only fundamental error in my suggestion was the location of tr.

Anyway, if you really want to correct the inconsistencies you need only edit the sources for the modules in question; the names which appear in /proc/modules appear to be defined in, for example, drivers/usb/host/uhci-hcd.c, where the .description section of the module is set. Or change the filenames, though I don't know how that will fly with the make process.

Signature · 2 lines
Ryan Reich
[email protected]
#9 lore raw
Quoted 7 lines · Ryan Reich

Anyway, if you really want to correct the inconsistencies you need only edit the sources for the modules in question; the names which appear in /proc/modules appear to be defined in, for example, drivers/usb/host/uhci-hcd.c, where the .description section of the module is set. Or change the filenames, though I don't know how that will fly with the make process.

Of course I could patch the kernel sources to remove the inconsistencies in the module names. But IMHO it is much more important to convince the kernel developers that this inconsistency is bad.

Regards

Harri

#10 lore raw
Quoted 7 lines · Harald Dunkel
Quoted 5 lines · Ryan Reich

Anyway, if you really want to correct the inconsistencies you need only edit the sources for the modules in question; the names which appear in /proc/modules appear to be defined in, for example, drivers/usb/host/uhci-hcd.c, where the .description section of the module is set. Or change the filenames, though I don't know how that will fly

Quoted 8 lines

with

Quoted 2 lines

the make process.

Of course I could patch the kernel sources to remove the inconsistencies in the module names. But IMHO it is much more important to convince the kernel developers that this inconsistency is bad.

When 2.7 opens I will try to find out if we can rename all victims. I can tweak kbuild to warn for modules using '-', so we in the end can get rid of this inconsistency.

Rusty - do you see any problems with this?

Sam

#11 lore raw
Quoted 6 lines

When 2.7 opens I will try to find out if we can rename all victims. I can tweak kbuild to warn for modules using '-', so we in the end can get rid of this inconsistency.

Rusty - do you see any problems with this?

well I'm not Rusty by a long shot, but I do see a problem with this, or at least a subtlety that needs taking into account: the pcmcia subsystem (I include cardmgr in that) has an internal declaration of the module name inside the module, and the filename of the module *HAS* to match that or things just won't work. ide-cs is the one at risk here; I've had to debug what turned out to be a mismatch during 2.4.2-era on this one and believe me, that's not funny.

#12 lore raw
In reply to Sam Ravnborg
Quoted 9 lines · Sam Ravnborg

When 2.7 opens I will try to find out if we can rename all victims. I can tweak kbuild to warn for modules using '-', so we in the end can get rid of this inconsistency.

Rusty - do you see any problems with this?

Sam

Any chance to get this warning for 2.6?

What would happen if a symbol filename is changed by replacing the '-' with '_'?

The module-init-tools wouldn't care. I don't know the internals of kudzu, but discover2 uses modprobe to load the modules. The internal workarounds in discover2 for the inconsistency would become obsolete.

Maybe the alsa stuff? There are many alsa modules with '-'.

Would it be easier to fix the output of /proc/modules than renaming all modules with '-'?

Of course I would be glad to help.

Regards

Harri