Re: [PATCH v2 3/4] printk/btrfs: Handle more message headers

[Petr Mladek <pmladek@suse.com>]

On Tue 2016-12-13 10:21:27, Geert Uytterhoeven wrote:
> Hi Petr,
> 
> On Wed, Nov 9, 2016 at 1:41 PM, Petr Mladek <pmladek@suse.com> wrote:
> > The commit 4bcc595ccd80decb4245096e ("printk: reinstate KERN_CONT for
> > printing continuation lines") allows to define more message headers
> > for a single message. The motivation is that continuous lines might
> > get mixed. Therefore it make sense to define the right log level
> > for every piece of a cont line.
> >
> > The current btrfs_printk() macros do not support continuous lines
> > at the moment. But better be prepared for a custom messages and
> > avoid potential "lvl" buffer overflow.
> >
> > This patch iterates over the entire message header. It is interested
> > only into the message level like the original code.
> >
> > This patch also introduces PRINTK_MAX_SINGLE_HEADER_LEN. Three bytes
> > are enough for the message level header at the moment. But it used to
> > be three, see the commit 04d2c8c83d0e3ac5f ("printk: convert the format
> > for KERN_<LEVEL> to a 2 byte pattern").
> >
> > Also I fixed the default ratelimit level. It looked very strange
> > when it was different from the default log level.
> >
> > Signed-off-by: Petr Mladek <pmladek@suse.com>
> > ---
> >  fs/btrfs/super.c       | 26 +++++++++++++++-----------
> >  include/linux/printk.h |  2 ++
> >  2 files changed, 17 insertions(+), 11 deletions(-)
> >
> > diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> > index 74ed5aae6cea..c083d84eaa32 100644
> > --- a/fs/btrfs/super.c
> > +++ b/fs/btrfs/super.c
> > @@ -202,27 +202,31 @@ void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function
> >  void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
> >  {
> >         struct super_block *sb = fs_info->sb;
> > -       char lvl[4];
> > +       char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1];
> >         struct va_format vaf;
> >         va_list args;
> > -       const char *type = logtypes[4];
> > +       const char *type = NULL;
> >         int kern_level;
> >         struct ratelimit_state *ratelimit;
> 
> warning: ‘ratelimit’ may be used uninitialized in this function
> 
> So this triggered my attention. It seems my gcc (4.1.2) is not smart enough
> to notice that ratelimit will be set to a default value if !type.
>
> Still, IMHO the code is too convoluted for the casual reader.

I did not see this with gcc 4.8.5. But I agree that it might
be confusing. Please, find a proposed fix below.


> >         vaf.fmt = fmt;
> > diff --git a/include/linux/printk.h b/include/linux/printk.h
> > index a0859e169bc3..afe8ccec1672 100644
> > --- a/include/linux/printk.h
> > +++ b/include/linux/printk.h
> > @@ -10,6 +10,8 @@
> >  extern const char linux_banner[];
> >  extern const char linux_proc_banner[];
> >
> > +#define PRINTK_MAX_SINGLE_HEADER_LEN 2
> 
> I think you want to use this definition instead of the hardcoded "2" in
> printk_skip_level():
> 
> | static inline const char *printk_skip_level(const char *buffer)
> | {
> |         if (printk_get_level(buffer))
> |                 return buffer + 2;
> 
> return buffer + PRINTK_MAX_SINGLE_HEADER_LEN;

I do not like hardcoded constants either. But this is with sync
with the hardcoded buffer[0] and buffer[1] in printk_get_level().

PRINTK_MAX_SINGLE_HEADER_LEN is meant as the maximum length in case
of some non-standard lengths in the future. We used 3-byte headers
in the past. The name is confusing for this use case. I probably
should have avoided _MAX_ in the name. I would leave this for now.
It is well localized...


Here is the proposed fix for the warning:


>From 5b5a8f7ff4f2d5c33316e600a29d45314ac47a16 Mon Sep 17 00:00:00 2001
From: Petr Mladek <pmladek@suse.com>
Date: Tue, 13 Dec 2016 13:12:56 +0100
Subject: [PATCH] btrfs: Better handle btrfs_printk() defaults
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The commit 262c5e86fec7cfd ("printk/btrfs: handle more message headers")
triggers:

    warning: ‘ratelimit’ may be used uninitialized in this function

with gcc (4.1.2) and probably many other versions. The code actually
is correct but a bit twisted. Let's make it more straightforward
and set the default values at the beginning.

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Petr Mladek <pmladek@suse.com>
---
 fs/btrfs/super.c | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)

diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 180f910339f4..3b713b6fcc26 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -202,12 +202,12 @@ void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function
 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 {
 	struct super_block *sb = fs_info->sb;
-	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1];
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 	struct va_format vaf;
 	va_list args;
-	const char *type = NULL;
 	int kern_level;
-	struct ratelimit_state *ratelimit;
+	const char *type = logtypes[4];
+	struct ratelimit_state *ratelimit = &printk_limits[4];
 
 	va_start(args, fmt);
 
@@ -223,12 +223,6 @@ void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 		fmt += size;
 	}
 
-	if (!type) {
-		*lvl = '\0';
-		type = logtypes[4];
-		ratelimit = &printk_limits[4];
-	}
-
 	vaf.fmt = fmt;
 	vaf.va = &args;
 
-- 
1.8.5.6