Debian Bug report logs - #547902
dash: 'read' builtin reads only the first character from pipes or /proc files

Reply or subscribe to this bug.

Display info messages

Message #5 received at [email protected] (full text, mbox, reply):

From: Corin <[email protected]>

To: Debian Bug Tracking System <[email protected]>

Subject: dash only reads the first character from pipe (bash reads whole line)

Date: Tue, 22 Sep 2009 15:49:40 +0200

Package: dash
Version: 0.5.5.1-2.3
Severity: important


The 'read' command seems to only read the first character and not the whole line.

Example / comparison with bash:

# /bin/dash
# read MAX </proc/sys/net/netfilter/nf_conntrack_max
# echo $MAX
2

# /bin/bash
# read MAX </proc/sys/net/netfilter/nf_conntrack_max
# echo $MAX
262144


-- System Information:
Debian Release: squeeze/sid
  APT prefers testing
  APT policy: (990, 'testing'), (500, 'stable')
Architecture: amd64 (x86_64)

Kernel: Linux 2.6.26-2-vserver-amd64 (SMP w/4 CPU cores)
Locale: LANG=de_DE.UTF-8, LC_CTYPE=de_DE.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash

Versions of packages dash depends on:
ii  debianutils                  3.2.1       Miscellaneous utilities specific t
ii  dpkg                         1.15.3.1+b1 Debian package management system
ii  libc6                        2.9-25      GNU C Library: Shared libraries

dash recommends no packages.

dash suggests no packages.

-- debconf information:
* dash/sh: true

Message #12 received at [email protected] (full text, mbox, reply):

From: Gerrit Pape <[email protected]>

To: Corin <[email protected]>, [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Tue, 29 Sep 2009 11:59:55 +0000

On Tue, Sep 22, 2009 at 03:49:40PM +0200, Corin wrote:
> The 'read' command seems to only read the first character and not the whole line.
> 
> Example / comparison with bash:
> 
> # /bin/dash
> # read MAX </proc/sys/net/netfilter/nf_conntrack_max
> # echo $MAX
> 2
> 
> # /bin/bash
> # read MAX </proc/sys/net/netfilter/nf_conntrack_max
> # echo $MAX
> 262144

Hi, this doesn't look like a bug in dash, but in the kernel's
implementation of the /proc/sys/net/ hierarchy.  If you copy the file to
/tmp/, it works just fine.

cat <<EOT >t.c
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

int main(int argc, char **argv) {
	char c;
	int fd = open(argv[1], O_RDONLY);
	if (fd == -1) return 1;
	for (;;) {
		int i = read(fd, &c, 1);
		if (i == 0) break;
		if (i == -1) return 2;
		if (write(1, &c, 1) == -1) return 3;
	}
	write(1, "\n", 1);
	return 0;
}
EOT
gcc t.c
./a.out /proc/sys/net/unix/max_dgram_qlen
cat /proc/sys/net/unix/max_dgram_qlen
cp /proc/sys/net/unix/max_dgram_qlen /tmp/
./a.out /tmp/max_dgram_qlen

Regards, Gerrit.

Message #17 received at [email protected] (full text, mbox, reply):

From: Netskin | Corin Langosch <[email protected]>

To: [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Thu, 08 Oct 2009 13:34:40 +0200

Hi,

I'm still not sure if it's a kernel bug/misbehavior, as bash works fine 
on the same system. And even if it's a kernel bug, shouldn't dash 
contain the same work-around as bash seems to have?

Fact is, when upgrading debian (and so dash becomes the default shell) 
some applications break because of this. One example are munin plugins, 
which is where I discovered the bug/ regression. Changing the shebang 
line in the munin plugins from /bin/sh (symlink to /bin/dash) to 
/bin/bash solves the problems.

Corin

Message #26 received at [email protected] (full text, mbox, reply):

From: Gerrit Pape <[email protected]>

To: [email protected]

Cc: [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Tue, 13 Oct 2009 11:13:01 +0000

tags 547902 + patch
quit

Hi, please see http://bugs.debian.org/547902

Here's a patch from Christian Hohnstaedt <[email protected]>

Thanks, Gerrit.

---
Testparameter:
for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19; do dd if=/proc/sys/net/ipv4/tcp_wmem bs=$i 2>/dev/null; done

for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19; do dd if=/proc/sys/net/ipv4/tcp_wmem skip=$i bs=1 2>/dev/null |wc -c ; done

diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 9270125..038df14 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -2117,17 +2117,16 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
 #define TMPBUFLEN 21
 	int *i, vleft, first=1, neg, val;
 	unsigned long lval;
-	size_t left, len;
+	size_t left, len, off;
 	
 	char buf[TMPBUFLEN], *p;
 	char __user *s = buffer;
 	
-	if (!tbl_data || !table->maxlen || !*lenp ||
-	    (*ppos && !write)) {
+	if (!tbl_data || !table->maxlen || !*lenp) {
 		*lenp = 0;
 		return 0;
 	}
-	
+	off = 0;
 	i = (int *) tbl_data;
 	vleft = table->maxlen / sizeof(*i);
 	left = *lenp;
@@ -2176,25 +2175,31 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
 			if (conv(&neg, &lval, i, 1, data))
 				break;
 		} else {
+			loff_t diff;
 			p = buf;
-			if (!first)
-				*p++ = '\t';
-	
 			if (conv(&neg, &lval, i, 0, data))
 				break;
 
-			sprintf(p, "%s%lu", neg ? "-" : "", lval);
-			len = strlen(buf);
-			if (len > left)
-				len = left;
-			if(copy_to_user(s, buf, len))
-				return -EFAULT;
-			left -= len;
-			s += len;
+			len = sprintf(p, "%s%s%lu", first ? "" : "\t",
+						neg ? "-" : "", lval);
+			diff = *ppos - off;
+			off += len;
+			if (diff > 0 && diff < len) {
+				p += diff;
+				len -= diff;
+			}
+			if (off > *ppos) {
+				if (len > left)
+					len = left;
+				if(copy_to_user(s, p, len))
+					return -EFAULT;
+				left -= len;
+				s += len;
+			}
 		}
 	}
 
-	if (!write && !first && left) {
+	if (!write && !first && left && off >= *ppos) {
 		if(put_user('\n', s))
 			return -EFAULT;
 		left--, s++;

Message #36 received at [email protected] (full text, mbox, reply):

From: Ben Hutchings <[email protected]>

To: Gerrit Pape <[email protected]>, [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Sat, 17 Oct 2009 01:36:03 +0100

[Message part 1 (text/plain, inline)]

On Tue, 2009-10-13 at 11:13 +0000, Gerrit Pape wrote:
> tags 547902 + patch
> quit
> 
> Hi, please see http://bugs.debian.org/547902
> 
> Here's a patch from Christian Hohnstaedt <[email protected]>
[...]

Let us know when this change is accepted upstream.

Ben.

-- 
Ben Hutchings
It is impossible to make anything foolproof because fools are so ingenious.

[signature.asc (application/pgp-signature, inline)]

Message #43 received at [email protected] (full text, mbox, reply):

From: Gerrit Pape <[email protected]>

To: [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Fri, 13 Nov 2009 11:00:26 +0000

On Sat, Oct 17, 2009 at 01:36:03AM +0100, Ben Hutchings wrote:
> On Tue, 2009-10-13 at 11:13 +0000, Gerrit Pape wrote:
> > tags 547902 + patch
> > quit
> > 
> > Hi, please see http://bugs.debian.org/547902
> > 
> > Here's a patch from Christian Hohnstaedt <[email protected]>
> [...]
> 
> Let us know when this change is accepted upstream.

Sorry, I neither forwarded it upstream, nor follow upstream development.

Regards, Gerrit.

Message #48 received at [email protected] (full text, mbox, reply):

From: Ben Hutchings <[email protected]>

To: Gerrit Pape <[email protected]>, [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Fri, 13 Nov 2009 12:34:10 +0000

[Message part 1 (text/plain, inline)]

On Fri, 2009-11-13 at 11:00 +0000, Gerrit Pape wrote:
> On Sat, Oct 17, 2009 at 01:36:03AM +0100, Ben Hutchings wrote:
> > On Tue, 2009-10-13 at 11:13 +0000, Gerrit Pape wrote:
> > > tags 547902 + patch
> > > quit
> > > 
> > > Hi, please see http://bugs.debian.org/547902
> > > 
> > > Here's a patch from Christian Hohnstaedt <[email protected]>
> > [...]
> > 
> > Let us know when this change is accepted upstream.
> 
> Sorry, I neither forwarded it upstream, nor follow upstream development.

Then we will not apply your patch.

Ben.

-- 
Ben Hutchings
The two most common things in the universe are hydrogen and stupidity.

[signature.asc (application/pgp-signature, inline)]

Message #53 received at [email protected] (full text, mbox, reply):

From: Gerrit Pape <[email protected]>

To: [email protected]

Subject: [[email protected]: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)]

Date: Fri, 13 Nov 2009 14:29:20 +0000

FYI.  Sorry, I can't help further.

Regards, Gerrit.


----- Forwarded message from Ben Hutchings <[email protected]> -----

From: Ben Hutchings <[email protected]>
To: Gerrit Pape <[email protected]>, [email protected]
Date: Fri, 13 Nov 2009 12:34:10 +0000
Subject: Re: Bug#547902: dash only reads the first character from pipe
 (bash reads whole line)

On Fri, 2009-11-13 at 11:00 +0000, Gerrit Pape wrote:
> On Sat, Oct 17, 2009 at 01:36:03AM +0100, Ben Hutchings wrote:
> > On Tue, 2009-10-13 at 11:13 +0000, Gerrit Pape wrote:
> > > tags 547902 + patch
> > > quit
> > > 
> > > Hi, please see http://bugs.debian.org/547902
> > > 
> > > Here's a patch from Christian Hohnstaedt <[email protected]>
> > [...]
> > 
> > Let us know when this change is accepted upstream.
> 
> Sorry, I neither forwarded it upstream, nor follow upstream development.

Then we will not apply your patch.

Ben.

-- 
Ben Hutchings
The two most common things in the universe are hydrogen and stupidity.



----- End forwarded message -----

Message #58 received at [email protected] (full text, mbox, reply):

From: Moritz Muehlenhoff <[email protected]>

To: Gerrit Pape <[email protected]>

Cc: [email protected], [email protected], [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Wed, 18 Nov 2009 22:13:48 +0100

On Tue, Oct 13, 2009 at 11:13:01AM +0000, Gerrit Pape wrote:
> tags 547902 + patch
> quit

[ Adding Christian to CC: ]

Christian, did you submit your patch upstream?

Cheers,
        Moritz

> Hi, please see http://bugs.debian.org/547902
> 
> Here's a patch from Christian Hohnstaedt <[email protected]>
> 
> Thanks, Gerrit.
>
> 
> ---
> Testparameter:
> for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19; do dd if=/proc/sys/net/ipv4/tcp_wmem bs=$i 2>/dev/null; done
> 
> for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19; do dd if=/proc/sys/net/ipv4/tcp_wmem skip=$i bs=1 2>/dev/null |wc -c ; done
> 
> diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> index 9270125..038df14 100644
> --- a/kernel/sysctl.c
> +++ b/kernel/sysctl.c
> @@ -2117,17 +2117,16 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
>  #define TMPBUFLEN 21
>  	int *i, vleft, first=1, neg, val;
>  	unsigned long lval;
> -	size_t left, len;
> +	size_t left, len, off;
>  	
>  	char buf[TMPBUFLEN], *p;
>  	char __user *s = buffer;
>  	
> -	if (!tbl_data || !table->maxlen || !*lenp ||
> -	    (*ppos && !write)) {
> +	if (!tbl_data || !table->maxlen || !*lenp) {
>  		*lenp = 0;
>  		return 0;
>  	}
> -	
> +	off = 0;
>  	i = (int *) tbl_data;
>  	vleft = table->maxlen / sizeof(*i);
>  	left = *lenp;
> @@ -2176,25 +2175,31 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
>  			if (conv(&neg, &lval, i, 1, data))
>  				break;
>  		} else {
> +			loff_t diff;
>  			p = buf;
> -			if (!first)
> -				*p++ = '\t';
> -	
>  			if (conv(&neg, &lval, i, 0, data))
>  				break;
>  
> -			sprintf(p, "%s%lu", neg ? "-" : "", lval);
> -			len = strlen(buf);
> -			if (len > left)
> -				len = left;
> -			if(copy_to_user(s, buf, len))
> -				return -EFAULT;
> -			left -= len;
> -			s += len;
> +			len = sprintf(p, "%s%s%lu", first ? "" : "\t",
> +						neg ? "-" : "", lval);
> +			diff = *ppos - off;
> +			off += len;
> +			if (diff > 0 && diff < len) {
> +				p += diff;
> +				len -= diff;
> +			}
> +			if (off > *ppos) {
> +				if (len > left)
> +					len = left;
> +				if(copy_to_user(s, p, len))
> +					return -EFAULT;
> +				left -= len;
> +				s += len;
> +			}
>  		}
>  	}
>  
> -	if (!write && !first && left) {
> +	if (!write && !first && left && off >= *ppos) {
>  		if(put_user('\n', s))
>  			return -EFAULT;
>  		left--, s++;
> 
> 
> 

Message #66 received at [email protected] (full text, mbox, reply):

From: Christian Hohnstaedt <[email protected]>

To: Moritz Muehlenhoff <[email protected]>

Cc: Gerrit Pape <[email protected]>, [email protected], [email protected], [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Thu, 19 Nov 2009 09:27:47 +0100

On Wed, Nov 18, 2009 at 10:13:48PM +0100, Moritz Muehlenhoff wrote:
> On Tue, Oct 13, 2009 at 11:13:01AM +0000, Gerrit Pape wrote:
> > tags 547902 + patch
> > quit
> 
> [ Adding Christian to CC: ]
> 
> Christian, did you submit your patch upstream?

No, not yet.

It still has the problem that reading the value(s) bytewise returns garbage
if the value(s) change during reading.

But I can try to send it to lkml and see what they say.
Shall I point to this bug-report and may I add a "Tested-by" line ?

	Christian

> 
> Cheers,
>         Moritz
> 
> > Hi, please see http://bugs.debian.org/547902
> > 
> > Here's a patch from Christian Hohnstaedt <[email protected]>
> > 
> > Thanks, Gerrit.
> >
> > 
> > ---
> > Testparameter:
> > for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19; do dd if=/proc/sys/net/ipv4/tcp_wmem bs=$i 2>/dev/null; done
> > 
> > for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19; do dd if=/proc/sys/net/ipv4/tcp_wmem skip=$i bs=1 2>/dev/null |wc -c ; done
> > 
> > diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> > index 9270125..038df14 100644
> > --- a/kernel/sysctl.c
> > +++ b/kernel/sysctl.c
> > @@ -2117,17 +2117,16 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
> >  #define TMPBUFLEN 21
> >  	int *i, vleft, first=1, neg, val;
> >  	unsigned long lval;
> > -	size_t left, len;
> > +	size_t left, len, off;
> >  	
> >  	char buf[TMPBUFLEN], *p;
> >  	char __user *s = buffer;
> >  	
> > -	if (!tbl_data || !table->maxlen || !*lenp ||
> > -	    (*ppos && !write)) {
> > +	if (!tbl_data || !table->maxlen || !*lenp) {
> >  		*lenp = 0;
> >  		return 0;
> >  	}
> > -	
> > +	off = 0;
> >  	i = (int *) tbl_data;
> >  	vleft = table->maxlen / sizeof(*i);
> >  	left = *lenp;
> > @@ -2176,25 +2175,31 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
> >  			if (conv(&neg, &lval, i, 1, data))
> >  				break;
> >  		} else {
> > +			loff_t diff;
> >  			p = buf;
> > -			if (!first)
> > -				*p++ = '\t';
> > -	
> >  			if (conv(&neg, &lval, i, 0, data))
> >  				break;
> >  
> > -			sprintf(p, "%s%lu", neg ? "-" : "", lval);
> > -			len = strlen(buf);
> > -			if (len > left)
> > -				len = left;
> > -			if(copy_to_user(s, buf, len))
> > -				return -EFAULT;
> > -			left -= len;
> > -			s += len;
> > +			len = sprintf(p, "%s%s%lu", first ? "" : "\t",
> > +						neg ? "-" : "", lval);
> > +			diff = *ppos - off;
> > +			off += len;
> > +			if (diff > 0 && diff < len) {
> > +				p += diff;
> > +				len -= diff;
> > +			}
> > +			if (off > *ppos) {
> > +				if (len > left)
> > +					len = left;
> > +				if(copy_to_user(s, p, len))
> > +					return -EFAULT;
> > +				left -= len;
> > +				s += len;
> > +			}
> >  		}
> >  	}
> >  
> > -	if (!write && !first && left) {
> > +	if (!write && !first && left && off >= *ppos) {
> >  		if(put_user('\n', s))
> >  			return -EFAULT;
> >  		left--, s++;
> > 
> > 
> > 
Christian Hohnstaedt

-- 
Christian Hohnstaedt / Project Manager Hardware and Manufacturing

Innominate Security Technologies AG / protecting industrial networks
tel: +49.30.921028.208 / fax: +49.30.921028.020
Rudower Chaussee 13, D-12489 Berlin / http://www.innominate.com

Register Court: AG Charlottenburg, HR B 81603
Management Board: Dirk Seewald
Chairman of the Supervisory Board: Volker Bibelhausen

Message #74 received at [email protected] (full text, mbox, reply):

From: Moritz Muehlenhoff <[email protected]>

To: Christian Hohnstaedt <[email protected]>

Cc: Gerrit Pape <[email protected]>, [email protected], [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Tue, 24 Nov 2009 22:16:06 +0100

On Thu, Nov 19, 2009 at 09:27:47AM +0100, Christian Hohnstaedt wrote:
> On Wed, Nov 18, 2009 at 10:13:48PM +0100, Moritz Muehlenhoff wrote:
> > On Tue, Oct 13, 2009 at 11:13:01AM +0000, Gerrit Pape wrote:
> > > tags 547902 + patch
> > > quit
> > 
> > [ Adding Christian to CC: ]
> > 
> > Christian, did you submit your patch upstream?
> 
> No, not yet.

Please keep this bug CCed if you submit it.
 
> It still has the problem that reading the value(s) bytewise returns garbage
> if the value(s) change during reading.
> 
> But I can try to send it to lkml and see what they say.
> Shall I point to this bug-report and may I add a "Tested-by" line ?

I didn't test it myself, but maybe Corin can give feedback on this.

Cheers,
        Moritz

Message #82 received at [email protected] (full text, mbox, reply):

From: Moritz Muehlenhoff <[email protected]>

To: Christian Hohnstaedt <[email protected]>

Cc: [email protected]

Subject: Re: Bug#547902: dash only reads the first character from pipe (bash reads whole line)

Date: Fri, 6 May 2011 21:15:45 +0200

On Tue, Nov 24, 2009 at 10:16:06PM +0100, Moritz Muehlenhoff wrote:
> On Thu, Nov 19, 2009 at 09:27:47AM +0100, Christian Hohnstaedt wrote:
> > On Wed, Nov 18, 2009 at 10:13:48PM +0100, Moritz Muehlenhoff wrote:
> > > On Tue, Oct 13, 2009 at 11:13:01AM +0000, Gerrit Pape wrote:
> > > > tags 547902 + patch
> > > > quit
> > > 
> > > [ Adding Christian to CC: ]
> > > 
> > > Christian, did you submit your patch upstream?
> > 
> > No, not yet.
> 
> Please keep this bug CCed if you submit it.

Closing due to lack of activity.

Cheers,
        Moritz

Message #92 received at [email protected] (full text, mbox, reply):

From: Jonathan Nieder <[email protected]>

To: [email protected]

Cc: Moritz Muehlenhoff <[email protected]>

Subject: Re: dash only reads the first character from pipe (bash reads whole line)

Date: Fri, 6 May 2011 14:26:07 -0500

reassign 547906 dash 0.5.6.1-1~exp2
reopen 547906
tags 547906 = patch
quit

Moritz Muehlenhoff wrote:

> Closing due to lack of activity.

I'll take it.  Will reassign back to linux-2.6 if there is progress.

Message #107 received at [email protected] (full text, mbox, reply):

From: Gioele Barabucci <[email protected]>

To: [email protected]

Subject: Re: Bug #547902: dash only reads the first character from pipe (bash reads whole line)

Date: Sat, 14 Nov 2015 13:53:04 +0100

Control: tags -1 upstream

This problem is known upstream, see 
<http://thread.gmane.org/gmane.comp.shells.dash/1174>.

The suggested workaround is

    MAX=$(cat /proc/.../file)

Regards,

--
Gioele Barabucci <[email protected]>

Message #128 received at [email protected] (full text, mbox, reply):

From: Herbert Xu <[email protected]>

To: [email protected]

Subject: Upstream fix coming soon

Date: Sun, 19 May 2024 18:08:47 +0800

This will be fixed upstream soon with

https://patchwork.kernel.org/project/dash/patch/fd52d05affd400c2c2f2ba238ff83555d11ca648.1716096097.git.herbert@gondor.apana.org.au/

Cheers,
-- 
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Send a report that this bug log contains spam.