Re: knfsd problem in late 2.1

Bill Hawes (
Thu, 21 May 1998 11:40:44 -0400 wrote:
> Ok, this patch fixes knfsd. It does the following:
> - Changes all internal interfaces to use kdev_t (including the
> nfs filehandle cookie).
> - Convert dev_t to kdev_t in all interfaces to usermode (nfsctl()).
> The only exception is in the nfsfh cookie that is passed for a mount
> request, but mountd should treat that as opaque anyways so it should
> be no problem.
> - Add some KERN_DEBUGs to printk()s. There are still some missing, and
> far too many of them too (currently when I kill all nfsds in my shutdown
> script I get half a screenfull of crap)
> - Make linux/sunrpc/sched.h compile without -D__KERNEL__ [needed by
> the user tools].

Hi Andi,

Thanks for going through and making the kdev_t changes.

With regards to the KERN_DEBUG changes, most of them are OK, but I don't
agree with changing the classification of a few of them. I often mark
unexpected error conditions with a ! at the end, and these should be
classified as KERN_ERR to make sure they get logged (and hopefully

> if (!fhp->fh_dverified) {
> - printk("fh_update: fh not verified!\n");
> + printk(KERN_DEBUG "fh_update: fh not verified!\n");
> goto out;

> if (!inode) {
> - printk("fh_update: %s/%s still negative!\n",
> + printk(KERN_DEBUG "fh_update: %s/%s still negative!\n",
> dentry->d_parent->, dentry->;

> fhp->fh_dverified = 0;
> if (!dentry->d_count) {
> - printk("fh_put: %s/%s has d_count 0!\n",
> + printk(KERN_DEBUG "fh_put: %s/%s has d_count 0!\n",
> dentry->d_parent->, dentry->;
> return;


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to