In message <Pine.LNX.4.64.0709261036140.3121@xxxxxxxxxxxxxxxxxxxxxxxxx>, Jan Engelhardt writes:
>
> On Sep 25 2007 23:09, Erez Zadok wrote:
> >--- a/fs/unionfs/commonfops.c
> >+++ b/fs/unionfs/commonfops.c
> >@@ -394,8 +394,8 @@ int unionfs_file_revalidate(struct file *file, bool willwrite)
> > if (willwrite && IS_WRITE_FLAG(file->f_flags) &&
> > !IS_WRITE_FLAG(unionfs_lower_file(file)->f_flags) &&
> > is_robranch(dentry)) {
> >- printk(KERN_DEBUG "unionfs: do delay copyup of \"%s\"\n",
> >- dentry->d_name.name);
> >+ dprintk(KERN_DEBUG "unionfs: do delay copyup of \"%s\"\n",
> >+ dentry->d_name.name);
>
> Don't we have pr_debug() for that?
Jan, what's the policy on people writing their own debugging printk systems.
I've looked at what other file systems do, and found out that it varies a
lot: some use a simple mapping wrapper from printk->dprintk, while others
define complex ways to have debugging levels/bitmaps where users can
selectively decide (say, by a mount option), what debugging output they'd
like to see or not. Here's a small sampling of what some file systems use:
9p/fid.c: P9_DPRINTK(P9_DEBUG_VFS, "fid %d dentry %s\n",
autofs/inode.c: DPRINTK(("autofs: shutting down\n"));
lockd/svcproc.c: dprintk("lockd: GRANTED_RES called\n");
ncpfs/dir.c: DDPRINTK("ncp_lookup_validate: result=%d\n", val);
nfs/client.c: dprintk("<-- nfs_free_client()\n");
nfsd/export.c: dprintk("found domain %s\n", buf);
partitions/efi.c: Dprintk("GPT my_lba incorrect: %lld != %lld\n",
Not much standardization there :-) but "dprintk" does appear to be the more
common use.
I wanted something simple to allow me to not printk something that's just
for informational/debugging purposes, but w/o having to #ifdef the printk in
question, or define a complex debugging-level printk system.
Now, looking at pr_debug (in linux/kernel.h), and indeed some filesystems
(e.g., affs and configfs) use it. But pr_debug is only active when #define
DEBUG is on (not CONFIG_DEBUG). I didn't see a config option that enable
DEBUG: is there one?
>From other file systems, it appears that the more common convention is for
filesystem XYZ to offer a config option CONFIG_DEBUG_XYZ, which makes its
debugging more self-contained and keeps the config option closer to where
the f/s itself is enabled.
Thanks,
Erez.
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
