Re: error on TTY read: no such file or directory

Re: error on TTY read: no such file or directory

[Borsenkow Andrej]

lordzork@lordzork.com wrote:

> In the gothic chambers of the underworld on Sat 29 Sep 2001 at 16:26 -0400,
> Borsenkow Andrej muttered darkly:
> 
> 
>>Could you send me the full strace. I promise nothing but this resembles 
>>something :)
>>
>>
> 
> certainly. here it is.
> 
> t.
> 
> 


open("/etc/zsh/zshenv", O_RDONLY|O_NOCTTY|O_LARGEFILE) = 3
...

open("/tmp/login-trace", O_WRONLY|O_CREAT|O_TRUNC|O_NOCTTY|O_LARGEFILE, 
0666) = 3
fcntl64(2, F_DUPFD, 10)                 = 13
close(2)                                = 0
dup2(3, 2)                              = 2
...
ioctl(2, SNDCTL_TMR_TIMEBASE, 0xbffff8d8) = -1 ENOTTY (Inappropriate 
ioctl for device)

fd 2 does not look like tty, does it? I presume you do exec > 
/tmp/login-trace very early in your /etc/zshenv

Still there is something weird going on:

write(10, "\r\33[m\33[m\33[m\33[J\n\33[1;30m[ \33[1;37mlc"..., 270) = 270
write(10, "\33[K", 3)                   = 3
read(10, "l", 1)                        = 1
write(10, "l", 1)                       = 1
read(10, "s", 1)                        = 1
write(10, "\10ls", 3)                   = 3
read(10, "syJ\10\10\0\0\0\0\0\0\0\0\240G\f\10\0\0\0\0\1\0\0\0\0\0"..., 
1) = 1025

Oops. Where does this last line comes from?

-andrej

Re: error on TTY read: no such file or directory

[Borsenkow Andrej]

lordzork@lordzork.com wrote:

>>open("/tmp/login-trace", O_WRONLY|O_CREAT|O_TRUNC|O_NOCTTY|O_LARGEFILE, 
>>0666) = 3
>>fcntl64(2, F_DUPFD, 10)                 = 13
>>close(2)                                = 0
>>dup2(3, 2)                              = 2
>>...
>>ioctl(2, SNDCTL_TMR_TIMEBASE, 0xbffff8d8) = -1 ENOTTY (Inappropriate 
>>ioctl for device)
>>
>>fd 2 does not look like tty, does it? I presume you do exec > 
>>/tmp/login-trace very early in your /etc/zshenv
>>


You noticed it, dd not you? Have you checked your zshenv for stderr 
redrection? (it was to be exec 2> /tmp/login-trace, of course).


>>Still there is something weird going on:
>>
>>write(10, "\r\33[m\33[m\33[m\33[J\n\33[1;30m[ \33[1;37mlc"..., 270) = 270
>>write(10, "\33[K", 3)                   = 3
>>read(10, "l", 1)                        = 1
>>write(10, "l", 1)                       = 1
>>read(10, "s", 1)                        = 1
>>write(10, "\10ls", 3)                   = 3
>>read(10, "syJ\10\10\0\0\0\0\0\0\0\0\240G\f\10\0\0\0\0\1\0\0\0\0\0"..., 
>>1) = 1025
>>
>>Oops. Where does this last line comes from?
>>
>>
> 
> that is  probably where it booted me. i typed ls once and it ran, then
> again and it froze after the first 'l'. 
> 


It was not my question. I do not understand where this data comes from 
and how can read one byte return 1025 bytes. It looks more like a 
contents of readdir() but I have no idea how can it appear on shell input.

-andrej

PS please keep further conversation on the list.

Re: error on TTY read: no such file or directory

[lordzork]

In the gothic chambers of the underworld on Sat 29 Sep 2001 at 17:56 -0400,
Borsenkow Andrej muttered darkly:

 
> You noticed it, dd not you? Have you checked your zshenv for stderr 
> redrection? (it was to be exec 2> /tmp/login-trace, of course).

yes, that's what i have:

    ((UID)) || {
        set -x
        exec 2>|/tmp/login-trace
    }

 
 
> It was not my question. I do not understand where this data comes from 
> and how can read one byte return 1025 bytes. It looks more like a 
> contents of readdir() but I have no idea how can it appear on shell input.

like i said, this is where it booted me so i'm not really sure where
that data comes from...

t.

-- 
4.0k -rw-rw-r--    1 zork     zork       17 Apr 4 17:38 .signature

Re: error on TTY read: no such file or directory

[Borsenkow Andrej]

lordzork@lordzork.com wrote:

> In the gothic chambers of the underworld on Sat 29 Sep 2001 at 17:56 -0400,
> Borsenkow Andrej muttered darkly:
> 
>  
> 
>>You noticed it, dd not you? Have you checked your zshenv for stderr 
>>redrection? (it was to be exec 2> /tmp/login-trace, of course).
>>
> 
> yes, that's what i have:
> 
>     ((UID)) || {
>         set -x
>         exec 2>|/tmp/login-trace
>     }
> 
>  


So if you remove this redirection? Does it work?

Re: error on TTY read: no such file or directory

[lordzork]

In the gothic chambers of the underworld on Sat 29 Sep 2001 at 18:19 -0400,
Borsenkow Andrej muttered darkly:

> lordzork@lordzork.com wrote:
> 
> >In the gothic chambers of the underworld on Sat 29 Sep 2001 at 17:56 -0400,
> >Borsenkow Andrej muttered darkly:
> >
> > 
> >
> >>You noticed it, dd not you? Have you checked your zshenv for stderr 
> >>redrection? (it was to be exec 2> /tmp/login-trace, of course).
> >>
> >
> >yes, that's what i have:
> >
> >    ((UID)) || {
> >        set -x
> >        exec 2>|/tmp/login-trace
> >    }
> >
> > 
> 
> 
> So if you remove this redirection? Does it work?
> 
> 
> 

no. that redirection was put there to help debug the problem, it makes
no difference whether it's there or not.

t.


-- 
4.0k -rw-rw-r--    1 zork     zork       17 Apr 4 17:38 .signature

Re: error on TTY read: no such file or directory

[Bart Schaefer]

On Sep 29,  6:11pm, lordzork@lordzork.com wrote:
}
}     ((UID)) || {
}         set -x
}         exec 2>|/tmp/login-trace
}     }

You can take that out now, it isn't helping.

} > It was not my question. I do not understand where this data comes from 
} > and how can read one byte return 1025 bytes. It looks more like a 
} > contents of readdir() but I have no idea how can it appear on shell input.

The data printed by strace is misleading.  read() is being called with
the address of single character (line 405 of zle_main.c):

	    r = read(SHTTY, &cc, 1);

Strace displays the arguments as they were passed *to* the read() call,
not as they are returned, so what you're seeing in the trace is just
stack garbage.

Nevertheless, the problem is that read() is returning a positive integer
greater than the number of bytes it was asked to read.  This confuses
the getkey() loop, which only expects that `r' can be 1, 0, or -1 at
that point.  You need to find out why your read() call is so badly broken!

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: error on TTY read: no such file or directory

[lordzork]

In the gothic chambers of the underworld on Sat 29 Sep 2001 at 18:39 -0400,
Bart Schaefer muttered darkly:

> On Sep 29,  6:11pm, lordzork@lordzork.com wrote:
> }
> }     ((UID)) || {
> }         set -x
> }         exec 2>|/tmp/login-trace
> }     }
> 
> You can take that out now, it isn't helping.
> 
> } > It was not my question. I do not understand where this data comes from 
> } > and how can read one byte return 1025 bytes. It looks more like a 
> } > contents of readdir() but I have no idea how can it appear on shell input.
> 
> The data printed by strace is misleading.  read() is being called with
> the address of single character (line 405 of zle_main.c):
> 
> 	    r = read(SHTTY, &cc, 1);
> 
> Strace displays the arguments as they were passed *to* the read() call,
> not as they are returned, so what you're seeing in the trace is just
> stack garbage.
> 
> Nevertheless, the problem is that read() is returning a positive integer
> greater than the number of bytes it was asked to read.  This confuses
> the getkey() loop, which only expects that `r' can be 1, 0, or -1 at
> that point.  You need to find out why your read() call is so badly broken!
> 

and this read() function is a part of what? zsh? glibc?

t.

-- 
4.0k -rw-rw-r--    1 zork     zork       17 Apr 4 17:38 .signature

Re: error on TTY read: no such file or directory

[Zefram]

lordzork@lordzork.com wrote:
>and this read() function is a part of what? zsh? glibc?

read() itself is in libc, but what strace shows you is the underlying
system call, which is part of the kernel.  The strace output you're
getting shows the kernel violating the rules of the read() API; but if
that may be broken, then it's also possible that ptrace() or strace is
broken instead.  What OS version is this?

-zefram

Re: error on TTY read: no such file or directory

[Bart Schaefer]

On Sep 29,  8:24pm, lordzork@lordzork.com wrote:
}
} In the gothic chambers of the underworld on Sat 29 Sep 2001 at 18:39 -0400,
} Bart Schaefer muttered darkly:
} 
} > You need to find out why your read() call is so badly broken!
} 
} and this read() function is a part of what? zsh? glibc?

It's a system call (that's why it shows up in strace output).

As a workaround you can try changing

	    if (r == 1)
		break;

to

	    if (r > 0)
		break;

at line 407 of Src/Zle/zle_main.c; but if the wrong number of bytes may
be returned by read(), all sorts of other things are going to break (not
just in zsh), so you shouldn't leave it at that.  You may have to take
this up on a linux kernel list.

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: error on TTY read: no such file or directory

[lordzork]

In the gothic chambers of the underworld on Sat 29 Sep 2001 at 21:24 -0400,
Bart Schaefer muttered darkly:

> It's a system call (that's why it shows up in strace output).
> 
> As a workaround you can try changing
> 
> 	    if (r == 1)
> 		break;
> 
> to
> 
> 	    if (r > 0)
> 		break;
> 
> at line 407 of Src/Zle/zle_main.c; but if the wrong number of bytes may
> be returned by read(), all sorts of other things are going to break (not
> just in zsh), so you shouldn't leave it at that.  You may have to take
> this up on a linux kernel list.
> 

i don't see how it can be the kernel, though. i've tried three different
kernel versions (2.4.7, 2.4.9, 2.4.10) and the problem remains with all
of them. furthermore, i've got virtually the exact setup duplicated on
two other boxes, with the same kernel versions. zsh works fine on both
boxes.

t. 

-- 
4.0k -rw-rw-r--    1 zork     zork       17 Apr 4 17:38 .signature

Re: error on TTY read: no such file or directory

[Bart Schaefer]

On Sep 30,  2:20am, Zefram wrote:
}
} The strace output you're getting shows the kernel violating the rules
} of the read() API; but if that may be broken, then it's also possible
} that ptrace() or strace is broken instead.

I don't think it's ptrace/strace that's broken, as the error output
that he's getting from zsh when not tracing is consistent with a > 1
return value from read().

I'd be suspicious that the kernel is expecting a 64-bit value as the
third argument to read() and is getting only a 32-bit value -- but in
that case I'd expect strace to show the whole 64 bits so that we'd be
able to tell how many bytes read() really was looking for; and that
would leave the question of where the other 1024 bytes came from, and
why stuffing 1025 bytes through the address of a single character on
the stack doesn't cause a more serious meltdown ...

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: error on TTY read: no such file or directory

[Borsenkow Andrej]

lordzork@lordzork.com wrote:

> In the gothic chambers of the underworld on Sat 29 Sep 2001 at 21:24 -0400,
> Bart Schaefer muttered darkly:
> 
> 
>>It's a system call (that's why it shows up in strace output).
>>
>>As a workaround you can try changing
>>
>>	    if (r == 1)
>>		break;
>>
>>to
>>
>>	    if (r > 0)
>>		break;
>>
>>at line 407 of Src/Zle/zle_main.c; but if the wrong number of bytes may
>>be returned by read(), all sorts of other things are going to break (not
>>just in zsh), so you shouldn't leave it at that.  You may have to take
>>this up on a linux kernel list.
>>
>>
> 
> i don't see how it can be the kernel, though. i've tried three different
> kernel versions (2.4.7, 2.4.9, 2.4.10) and the problem remains with all
> of them. furthermore, i've got virtually the exact setup duplicated on
> two other boxes, with the same kernel versions. zsh works fine on both
> boxes.
> 


What about glibc versions? Are they the same on all systems?

Re: error on TTY read: no such file or directory

[lordzork]

In the gothic chambers of the underworld on Sun 30 Sep 2001 at 02:25 -0400,
Borsenkow Andrej muttered darkly:

 
> What about glibc versions? Are they the same on all systems?
> 

yes, glibc 2.2.3 

-- 
4.0k -rw-rw-r--    1 zork     zork       17 Apr 4 17:38 .signature

Re: error on TTY read: no such file or directory

[lordzork]

i made a new strace, this time without zsh reading any of its start up
files. my idea was that this would make it easier to read, but the new
strace looks completely different from the old one, at least to me.

http://lzi.n3.net/~zork/zle_error/strace.txt

t.

ps - could people please just send replies to the list instead of also
sending a copy to my personal address? thanks.

-- 
4.0k -rw-rw-r--    1 zork     zork       17 Apr 4 17:38 .signature

Re: error on TTY read: no such file or directory

[Zefram]

lordzork@lordzork.com wrote:
>i made a new strace, this time without zsh reading any of its start up
>files. my idea was that this would make it easier to read, but the new
>strace looks completely different from the old one, at least to me.

It's still doing the same impossible thing.

-zefram

Re: error on TTY read: no such file or directory

[Borsenkow Andrej]

Zefram wrote:

> lordzork@lordzork.com wrote:
> 
>>i made a new strace, this time without zsh reading any of its start up
>>files. my idea was that this would make it easier to read, but the new
>>strace looks completely different from the old one, at least to me.
>>
> 
> It's still doing the same impossible thing.
> 


Have you built zsh on this system or used binaries from other? Now, we 
start to make all sort of wild guesses ...

Re: error on TTY read: no such file or directory

[Bart Schaefer]

On Sep 30,  1:38pm, Zefram wrote:
} Subject: Re: error on TTY read: no such file or directory
}
} lordzork@lordzork.com wrote:
} >i made a new strace, this time without zsh reading any of its start up
} >files. my idea was that this would make it easier to read, but the new
} >strace looks completely different from the old one, at least to me.
} 
} It's still doing the same impossible thing.

I never saw the previous trace.   Looking at this one, I note that it's
doing this impossible thing a lot, but only on fd 10, and sometimes zsh
just ignores the problem and loops.  (That would happen when errno == 0,
or if the impossible occurred somewhere other than in getkey().)

Ignoring the problem would mean that zsh discarded at least one char of
input if there really were any, but there's no evidence of input being
lost.  That in turn makes me think that reads from the tty are behaving
as if they're non-blocking, but for some reason are returning 1025 when
they should be returning -1 with EWOULDBLOCK.

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: error on TTY read: no such file or directory

[Borsenkow Andrej]

Bart Schaefer wrote:

> On Sep 30,  1:38pm, Zefram wrote:
> } Subject: Re: error on TTY read: no such file or directory
> }
> } lordzork@lordzork.com wrote:
> } >i made a new strace, this time without zsh reading any of its start up
> } >files. my idea was that this would make it easier to read, but the new
> } >strace looks completely different from the old one, at least to me.
> } 
> } It's still doing the same impossible thing.
> 
> I never saw the previous trace.   Looking at this one, I note that it's
> doing this impossible thing a lot, but only on fd 10, and sometimes zsh
> just ignores the problem and loops.  (That would happen when errno == 0,
> or if the impossible occurred somewhere other than in getkey().)
> 


In first strace I got it does it exactly once and then bails out. Not 
that it helps much.


> Ignoring the problem would mean that zsh discarded at least one char of
> input if there really were any, but there's no evidence of input being
> lost.  That in turn makes me think that reads from the tty are behaving
> as if they're non-blocking, but for some reason are returning 1025 when
> they should be returning -1 with EWOULDBLOCK.
> 


I very much like to see what happens with "standard" mingetty instead of 
agetty. There is always a small chance it sets tty into some weird mode.


-andrej

Re: error on TTY read: no such file or directory

[Peter Stephenson]

Bart Schaefer wrote:
> I'd be suspicious that the kernel is expecting a 64-bit value as the
> third argument to read() and is getting only a 32-bit value -- but in
> that case I'd expect strace to show the whole 64 bits so that we'd be
> able to tell how many bytes read() really was looking for; and that
> would leave the question of where the other 1024 bytes came from, and
> why stuffing 1025 bytes through the address of a single character on
> the stack doesn't cause a more serious meltdown ...

Or could it be the other way around and _LARGEFILE_SOURCE and the
corresponding definitions, such as open64, aren't being handled correctly?
It's certainly at least partially in effect because of the stat64.

May be a complete red herring.

-- 
Peter Stephenson <pws@csr.com>                  Software Engineer
CSR Ltd., Science Park, Milton Road,
Cambridge, CB4 0WH, UK                          Tel: +44 (0)1223 392070


**********************************************************************
The information transmitted is intended only for the person or
entity to which it is addressed and may contain confidential 
and/or privileged material. 
Any review, retransmission, dissemination or other use of, or
taking of any action in reliance upon, this information by 
persons or entities other than the intended recipient is 
prohibited.  
If you received this in error, please contact the sender and 
delete the material from any computer.
**********************************************************************