The attatched patch fixes smbfs compilation in 2.5. The problem was that the
paranoia macro required a format string, and one or more substitution elements.
4 callers only used the string with no substitution. This patch defines a
paranoia2 which allows for not substitutions ;)
Compilation tested only, as i can not run latest 2.5 due to raid5 not
compiling, but this has to be ok, since it doesn't change any code paths.
Erik McKee
# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher.
# This patch includes the following deltas:
# ChangeSet 1.484 -> 1.485
# fs/smbfs/smb_debug.h 1.2 -> 1.3
# fs/smbfs/proc.c 1.20 -> 1.21
# fs/smbfs/sock.c 1.7 -> 1.8
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 02/06/15 [email protected] 1.485
# sock.c:
# Same edit as proc.c
# proc.c:
# This makes use of the PARANOIA2 to allow compilation
# smb_debug.h:
# Added PARANOIA2 to handle cases where there are no
# format specifiers in the string
# --------------------------------------------
#
diff -Nru a/fs/smbfs/proc.c b/fs/smbfs/proc.c
--- a/fs/smbfs/proc.c Sat Jun 15 23:12:04 2002
+++ b/fs/smbfs/proc.c Sat Jun 15 23:12:05 2002
@@ -872,7 +872,7 @@
goto out;
}
if (smb_valid_packet(s->packet) != 0) {
- PARANOIA("invalid packet!\n");
+ PARANOIA2("invalid packet!\n");
goto out;
}
diff -Nru a/fs/smbfs/smb_debug.h b/fs/smbfs/smb_debug.h
--- a/fs/smbfs/smb_debug.h Sat Jun 15 23:12:04 2002
+++ b/fs/smbfs/smb_debug.h Sat Jun 15 23:12:04 2002
@@ -12,8 +12,10 @@
*/
#ifdef SMBFS_PARANOIA
# define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__, ## a)
+# define PARANOIA2(f) printk(KERN_NOTICE "%s: "f, __FUNCTION__)
#else
# define PARANOIA(f, a...) do { ; } while(0)
+# defind PARANOIA2(f) do { ; } while(0)
#endif
/* lots of debug messages */
diff -Nru a/fs/smbfs/sock.c b/fs/smbfs/sock.c
--- a/fs/smbfs/sock.c Sat Jun 15 23:12:05 2002
+++ b/fs/smbfs/sock.c Sat Jun 15 23:12:05 2002
@@ -125,7 +125,7 @@
result = -EIO;
if (job->sk->dead) {
- PARANOIA("sock dead!\n");
+ PARANOIA2("sock dead!\n");
break;
}
@@ -189,7 +189,7 @@
{
#ifdef SMBFS_PARANOIA
if (!smb_valid_socket(file->f_dentry->d_inode))
- PARANOIA("bad socket!\n");
+ PARANOIA2("bad socket!\n");
#endif
return SOCKET_I(file->f_dentry->d_inode);
}
@@ -299,7 +299,7 @@
VERBOSE("closing socket %p\n", server_sock(server));
#ifdef SMBFS_PARANOIA
if (server_sock(server)->sk->data_ready == smb_data_ready)
- PARANOIA("still catching keepalives!\n");
+ PARANOIA2("still catching keepalives!\n");
#endif
server->sock_file = NULL;
fput(file);
@@ -585,7 +585,7 @@
return result;
out_no_mem:
- PARANOIA("couldn't allocate data area\n");
+ PARANOIA2("couldn't allocate data area\n");
result = -ENOMEM;
goto out;
out_too_long:
__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com
THis is from the bk tree. It's gcc 2.95.3. That solution might be a bettr one
after all ;) However, would the stringifying done here to get the function
name in there mess that up?
Erik McKee
--- Urban Widmark <[email protected]> wrote:
> On Sat, 15 Jun 2002, Erik McKee wrote:
>
> > diff -Nru a/fs/smbfs/smb_debug.h b/fs/smbfs/smb_debug.h
> > --- a/fs/smbfs/smb_debug.h Sat Jun 15 23:12:04 2002
> > +++ b/fs/smbfs/smb_debug.h Sat Jun 15 23:12:04 2002
> > @@ -12,8 +12,10 @@
> > */
> > #ifdef SMBFS_PARANOIA
> > # define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__, ##
> a)
> > +# define PARANOIA2(f) printk(KERN_NOTICE "%s: "f, __FUNCTION__)
>
> Are you looking at BK, the 2.5.21 tree I'm looking at still has:
> #define PARANOIA(x...) printk(KERN_NOTICE __FUNCTION__ ": " x)
>
> I assume you are using gcc 3.x? (which one?)
> I don't get any warnings/errors on 2.96.
>
>
> I think having two macros for exactly the same thing is ugly. A better
> solution might be to borrow some code from arch/ia64/kernel/unaligned.c
>
> #define PARANOIA(f...) \
> do { printk(KERN_NOTICE "%s: ", __FUNCTION__); printk(f); } while(0)
>
> (untested, and I think this will print an extra "<4>" ?)
> Unless someone has a better idea. In any case, the other printk macros in
> smb_debug.h needs treatment too and not just the PARANOIA macro.
>
>
> If you are cleaning things up I think that the following also sometimes
> use debug macros with a single string, but with a 'macro(a, b...)' syntax:
>
> drivers/net/pci-skeleton.c
> drivers/char/i810_rng.c
> sound/oss/via82cxxx_audio.c
> DPRINTK
>
> drivers/hotplug/pci_hotplug_{core,util}.c
> dbg/err/info/warn
>
> drivers/char/machzwd.c
> dprnintk
>
> drivers/ieee1394/sbp2.c
> SBP2_ORB_DEBUG
>
> net/ipv4/netfilter/ipt_ULOG.c
> net/ipv4/netfilter/ip_conntrack_irc.c
> DEBUGP
>
> sound/oss/vwsnd.c
> DBGP
>
> fs/ntfs/debug.h
> ntfs_warning/ntfs_error
>
>
> Many of them are not enabled by default, and maybe they have already been
> taken care of.
>
> And maybe run suggested changes by the respective maintainers first.
> Thanks.
>
> /Urban
>
__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com
On Sun, 16 Jun 2002, Erik McKee wrote:
> THis is from the bk tree. It's gcc 2.95.3. That solution might be a bettr one
> after all ;) However, would the stringifying done here to get the function
> name in there mess that up?
stringifying?
The whole point of the original change was to not do any string
concatenation of __FUNCTION__, but it is a string already.
However, could you try this change instead:
-# define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__, ## a)
+# define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__ , ## a)
Apparently an extra space before the comma before ## is supposed to
matter. Several of the macros I pointed at before already do that.
See also
http://gcc.gnu.org/onlinedocs/gcc-3.1/cpp/Variadic-Macros.html
/Urban
On Sun, 16 Jun 2002, Urban Widmark wrote:
>
> > # define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__, ## a)
> > +# define PARANOIA2(f) printk(KERN_NOTICE "%s: "f, __FUNCTION__)
>
> Are you looking at BK, the 2.5.21 tree I'm looking at still has:
> #define PARANOIA(x...) printk(KERN_NOTICE __FUNCTION__ ": " x)
Should not
# define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__ , ## a)
work? Notice the one added extra space between __FUNCTION__ and ",".
Linus
On Sun, 16 Jun 2002, Linus Torvalds wrote:
> Should not
>
> # define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__ , ## a)
>
> work? Notice the one added extra space between __FUNCTION__ and ",".
That's what the gcc manual says.
There was a separate mail where I dropped some people I figured didn't
care too much about smbfs debug macros from the Cc list.
I will get back to you when someone with 2.95.x has verified it - or if
you know that it is supposed to work just change it. It's in use in
other places in the kernel tree.
/Urban
On Sat, 15 Jun 2002, Erik McKee wrote:
> diff -Nru a/fs/smbfs/smb_debug.h b/fs/smbfs/smb_debug.h
> --- a/fs/smbfs/smb_debug.h Sat Jun 15 23:12:04 2002
> +++ b/fs/smbfs/smb_debug.h Sat Jun 15 23:12:04 2002
> @@ -12,8 +12,10 @@
> */
> #ifdef SMBFS_PARANOIA
> # define PARANOIA(f, a...) printk(KERN_NOTICE "%s: " f, __FUNCTION__, ## a)
> +# define PARANOIA2(f) printk(KERN_NOTICE "%s: "f, __FUNCTION__)
Are you looking at BK, the 2.5.21 tree I'm looking at still has:
#define PARANOIA(x...) printk(KERN_NOTICE __FUNCTION__ ": " x)
I assume you are using gcc 3.x? (which one?)
I don't get any warnings/errors on 2.96.
I think having two macros for exactly the same thing is ugly. A better
solution might be to borrow some code from arch/ia64/kernel/unaligned.c
#define PARANOIA(f...) \
do { printk(KERN_NOTICE "%s: ", __FUNCTION__); printk(f); } while(0)
(untested, and I think this will print an extra "<4>" ?)
Unless someone has a better idea. In any case, the other printk macros in
smb_debug.h needs treatment too and not just the PARANOIA macro.
If you are cleaning things up I think that the following also sometimes
use debug macros with a single string, but with a 'macro(a, b...)' syntax:
drivers/net/pci-skeleton.c
drivers/char/i810_rng.c
sound/oss/via82cxxx_audio.c
DPRINTK
drivers/hotplug/pci_hotplug_{core,util}.c
dbg/err/info/warn
drivers/char/machzwd.c
dprnintk
drivers/ieee1394/sbp2.c
SBP2_ORB_DEBUG
net/ipv4/netfilter/ipt_ULOG.c
net/ipv4/netfilter/ip_conntrack_irc.c
DEBUGP
sound/oss/vwsnd.c
DBGP
fs/ntfs/debug.h
ntfs_warning/ntfs_error
Many of them are not enabled by default, and maybe they have already been
taken care of.
And maybe run suggested changes by the respective maintainers first.
Thanks.
/Urban