Gilles Espinasse | 15 Feb 06:43 2010
Picon

http://bugzilla.kerneltools.org/ display uninterpreted perl script

Everything is in the title

Gilles
Gilles Espinasse | 16 Feb 15:02 2010
Picon

[PATCH] Fix depmod -e without -F or -E unterminated warning

Signed-off-by: Gilles Espinasse <g.esp@...>
---
 depmod.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/depmod.c b/depmod.c
index dfd210d..b739fa9 100644
--- a/depmod.c
+++ b/depmod.c
 <at>  <at>  -1342,7 +1342,7  <at>  <at>  int main(int argc, char *argv[])
 	else if (system_map)
 		load_system_map(system_map);
 	else if (print_unknown) {
-		warn("-e needs -E or -F");
+		warn("-e needs -E or -F\n");
 		print_unknown = 0;
 	}

--

-- 
1.6.4.4

Jon Masters | 16 Feb 17:05 2010
Picon

Re: [PATCH] Fix depmod -e without -F or -E unterminated warning

On Tue, 2010-02-16 at 15:02 +0100, Gilles Espinasse wrote:
> Signed-off-by: Gilles Espinasse <g.esp@...>
> ---
>  depmod.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/depmod.c b/depmod.c
> index dfd210d..b739fa9 100644
> --- a/depmod.c
> +++ b/depmod.c
>  <at>  <at>  -1342,7 +1342,7  <at>  <at>  int main(int argc, char *argv[])
>  	else if (system_map)
>  		load_system_map(system_map);
>  	else if (print_unknown) {
> -		warn("-e needs -E or -F");
> +		warn("-e needs -E or -F\n");
>  		print_unknown = 0;
>  	}

Thanks, I am testing a few other patches at the moment and plan a new
tree on Friday. Then we get back into regular Friday updates. I meant to
push that last week but got distracted with that netfilter ugly bug.

Jon.

Jon Masters | 23 Feb 06:57 2010
Picon

Re: [PATCH] Fix depmod -e without -F or -E unterminated warning

On Tue, 2010-02-16 at 11:05 -0500, Jon Masters wrote:
> On Tue, 2010-02-16 at 15:02 +0100, Gilles Espinasse wrote:
> > Signed-off-by: Gilles Espinasse <g.esp@...>
> > ---
> >  depmod.c |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> > 
> > diff --git a/depmod.c b/depmod.c
> > index dfd210d..b739fa9 100644
> > --- a/depmod.c
> > +++ b/depmod.c
> >  <at>  <at>  -1342,7 +1342,7  <at>  <at>  int main(int argc, char *argv[])
> >  	else if (system_map)
> >  		load_system_map(system_map);
> >  	else if (print_unknown) {
> > -		warn("-e needs -E or -F");
> > +		warn("-e needs -E or -F\n");
> >  		print_unknown = 0;
> >  	}
> 
> Thanks, I am testing a few other patches at the moment and plan a new
> tree on Friday. Then we get back into regular Friday updates. I meant to
> push that last week but got distracted with that netfilter ugly bug.

Should be out later today. I'm working on some patches from Miloslav
Trmac at the moment (modprobe whitelist) and will then do a push.

Jon.

(Continue reading)

Jon Masters | 24 Feb 21:40 2010

Re: [PATCH] modprobe does not print out dependee module falures.

On Fri, 2010-01-29 at 23:01 +0000, Matthew W. S. Bell wrote:
> When a failure occurs loading a dependee module, an error for this is
> not printed, but an incorrect error is printed for each dependent
> module. This patch changes this.

I see what you're trying to solve here, but the existing code has to
handle the case of loading multiple modules by means of dependencies
*and* in the case that -a is specified on the command line. I am
thinking the following:

-               if ((error != warn) || (verbose))
+               if (!all || verbose)
                        error("Error inserting %s (%s): %s\n",
                              mod->modname, mod->filename,
                              insert_moderror(errno));

Needs to be:

if ((!all && (error != warn)) || (verbose))

Jon.

Alan Jenkins | 25 Feb 18:11 2010
Picon

Queued cleanups

Thanks for pulling the forkbomb-prevention work.  Here's a re-post of 
the memory leak patches.

Next on my list was a prototype I started to make Andreas' softdep code 
a bit simpler.  It reads all the config file data into memory at the 
start, instead of re-parsing it for each module.  (I think it freed it 
all afterwards, but it could save the effort and just leave the pointers 
around so it doesn't look like a leak to valgrind).  The idea was to 
avoid having to pass all the filenames to insmod() (which was needed in 
case it called do_softdep()).

---
The following changes since commit d64f843b5aeec65a725b7eed267b6a32cad190a2:
  Jon Masters (1):
        pre-release: initial version for 3.12

are available in the git repository at:

  git@...:sourcejedi/module-init-tools.git master

Alan Jenkins (6):
      modprobe: clean up mit_ignore_loaded flag in rmmod()
      modprobe: simplify insmod() calling convention - don't free optstring
      modprobe: fix potential memory leak on failure path
      elfops: fix crash on grab_fd() failure
      modprobe: fix memory leak when built with zlib support
      insmod: fix memory leak

 elfops.c      |   39 +++++------------------
 elfops.h      |    1 -
(Continue reading)

Jon Masters | 25 Feb 18:31 2010
Picon

Re: Queued cleanups

On Thu, 2010-02-25 at 17:11 +0000, Alan Jenkins wrote:

> Thanks for pulling the forkbomb-prevention work.  Here's a re-post of 
> the memory leak patches.

Cool. Note that there are now several branches in my tree (or will be
when I push the latest stuff if not already on master):

* master - the head of the tree
* dev - development bits
* topic branches - new features have a topic

Please post against master as ususal. I finally learned more of the
useful workflow around git and now have all of your trees as remotes.
It's amazing how much I should have invested in figuring out git ;)

FWIW. I am currently looking at:

*). Some whitelist patches proposed via Fedora.
*). A regression due to incorrect blocking behavior.
*). A crash caused by the VMWare modules due to bad modules.
*). A few other abrt related crashes to mostly the above.
*). A problem with full filesystems.

I expect to have most of these finished this week too.

> Next on my list was a prototype I started to make Andreas' softdep code 
> a bit simpler.  It reads all the config file data into memory at the 
> start, instead of re-parsing it for each module.  (I think it freed it 
> all afterwards, but it could save the effort and just leave the pointers 
(Continue reading)

Matthew W. S. Bell | 26 Feb 02:11 2010
Picon

Re: [PATCH] modprobe does not print out dependee module falures.

On Wed, 2010-02-24 at 15:40 -0500, Jon Masters wrote:
> On Fri, 2010-01-29 at 23:01 +0000, Matthew W. S. Bell wrote:
> > When a failure occurs loading a dependee module, an error for this is
> > not printed, but an incorrect error is printed for each dependent
> > module. This patch changes this.
> 
> I see what you're trying to solve here, but the existing code has to
> handle the case of loading multiple modules by means of dependencies
> *and* in the case that -a is specified on the command line. I am
> thinking the following:
> 
> -               if ((error != warn) || (verbose))
> +               if (!all || verbose)
>                         error("Error inserting %s (%s): %s\n",
>                               mod->modname, mod->filename,
>                               insert_moderror(errno));
> 
> Needs to be:
> 
> if ((!all && (error != warn)) || (verbose))

Er, this reintroduces the bug again:

->: depends

!all: m->m->m->fail (no actual error message, but dependency errors)
      fail (error message)

 all: m->m->m->fail (no actual error message, but dependency errors),
      fail (no error message)
(Continue reading)

Jon Masters | 2 Mar 11:34 2010
Picon

Re: Queued cleanups

On Thu, 2010-02-25 at 17:11 +0000, Alan Jenkins wrote:

> Next on my list was a prototype I started to make Andreas' softdep code 
> a bit simpler.

Don't worry too much about my other mail, I see the three softdep
branches in your tree - I fetched them to take a look. More doc cleanups
in the dev branch now, and a fix for the locking problem coming shortly.

Jon.

Alan Jenkins | 2 Mar 11:42 2010
Picon

Re: Queued cleanups

Jon Masters wrote:
> On Thu, 2010-02-25 at 12:31 -0500, Jon Masters wrote:
>
>   
>> *). A crash caused by the VMWare modules due to bad modules.
>> *). A few other abrt related crashes to mostly the above.
>>     
>
> These are due to the previously discovered problem with recently deleted
> modules that was fixed in the previous release. My "dev" branch
> currently has a few things building up. I pushed some doc fixes earlier
> and have some more stuff pending to push later today. Out of interest,
> where do you stand on the softdep stuff at the moment? I would like to
> find some time to poke this week.
>
> Jon.
>   

I've completed it now, although it turns out I mis-remembered what the 
cleanup was actually for.  It doesn't reduce complexity much; what it 
does is remove the remaining memory leaks.

It actually ends with more lines of code.  Much of the extra code is in 
read_aliases().  For the plain-text modules.alias file, it's not ideal 
to re-use parse_config_file() and then worry about non-alias lines.  So 
I added a separate parser for modules.alias.

You don't necessarily need to take "softdep-merge" as-is.  I did the 
merge because my leak fixes were split between "cleanups" and 
"softdep-cleanups".  I had to merge them in order to test that I had 
(Continue reading)


Gmane