quotactl

quotactl

[Francis Lau]

Hi,

For some reason, quotactl does not work for me even though I have QUOTA 
turned on in the kernel.

 struct mem_dqblk mydq;
 i = quotactl (QCMD(Q_GETQUOTA,USRQUOTA), "/dev/md5", 2099, &mydq);

i always returns -1 with errno = 22 (EINVAL The kernel has not 
been compiled with the QUOTA  option. cmd is invalid.)

Has anyone seen this before? Any suggestions are welcome.

Thanks,
Francis

Re: quotactl

[Chris Nanakos]

make a clean compilation of the kernel from the beggining.(use make clean &&
make mrproper).Be sure that no object files exist when you start building
your new kernel.Try this....and e-mail again:)



Best regards,
Chris.

Re: quotactl

[Francis Lau]

> make a clean compilation of the kernel from the beggining.(use make clean &&
> make mrproper).Be sure that no object files exist when you start building
> your new kernel.Try this....and e-mail again:)
> 
I'm running RedHat 9 with the stock 2.4.20 SMP kernel.  Unfortunately, I 
cannot (not allowed to) recompile the kernel...Any other suggestions?

Thanks
Francis

Re: quotactl

[Darío Mariani]

   If you did not recompile the kernel, it has quota support. Has the FS 
you are quering been mounted with the usrquota option?

Francis Lau wrote:
>>make a clean compilation of the kernel from the beggining.(use make clean &&
>>make mrproper).Be sure that no object files exist when you start building
>>your new kernel.Try this....and e-mail again:)
>>
> 
> I'm running RedHat 9 with the stock 2.4.20 SMP kernel.  Unfortunately, I 
> cannot (not allowed to) recompile the kernel...Any other suggestions?
> 
> Thanks
> Francis
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: quotactl

[Francis Lau]

On Wed, 2 Jul 2003, Darío Mariani wrote:

>    If you did not recompile the kernel, it has quota support. Has the FS 
> you are quering been mounted with the usrquota option?

Yes the FS is mounted with the usrqutoa option.  I can use repquota, 
edquota, quota, quotacheck, quotaon, quotaoff...etc. with no problems.  
I only have problems with I use quotactl in my C code.  

Francis

> 
> Francis Lau wrote:
> >>make a clean compilation of the kernel from the beggining.(use make clean &&
> >>make mrproper).Be sure that no object files exist when you start building
> >>your new kernel.Try this....and e-mail again:)
> >>
> > 
> > I'm running RedHat 9 with the stock 2.4.20 SMP kernel.  Unfortunately, I 
> > cannot (not allowed to) recompile the kernel...Any other suggestions?
> > 
> > Thanks
> > Francis
> > 
> > -
> > To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > 
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

-
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

mid file deletion

[John T. Williams]

What is the best way to remove data from the middle of a file. Here is the
Scenario: I have a file that is 20MB in size, about ¼ th way through it is a
1 line I want to remove. I can use lseek to bring me to the beginning of the
section I want to remove, and then lseek again to bring me to the end of
that section. The only thing I can think to do is load everything from the
file after the point I want to delete into memory, remove the part I don't
want from what is stored in memory (or simply don't load it), truncate the
file at the point where the part I wanted to delete began, and re-write the
rest of the file. The alternative (since 20MB is rather big for loading into
memory on a multi-User multi-Media machine) is to load the bits I want to
keep into a temporary file, rather into memory. However, I was thinking that
for removing 20bytes of information, reading and writing 15MB to 30MB is a
lot of work. Is there a better way?




To give this more of a real world perspective. I'm downloading e-mails into
a mbox file and I'm trying to figure out how to delete 1 email from a mbox
file that could potentially store 1,000's of e-mails an any given time.







-
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: mid file deletion

[Glynn Clements]

John T. Williams wrote:

> What is the best way to remove data from the middle of a file. Here is the
> Scenario: I have a file that is 20MB in size, about ¼ th way through it is a
> 1 line I want to remove. I can use lseek to bring me to the beginning of the
> section I want to remove, and then lseek again to bring me to the end of
> that section. The only thing I can think to do is load everything from the
> file after the point I want to delete into memory, remove the part I don't
> want from what is stored in memory (or simply don't load it), truncate the
> file at the point where the part I wanted to delete began, and re-write the
> rest of the file. The alternative (since 20MB is rather big for loading into
> memory on a multi-User multi-Media machine) is to load the bits I want to
> keep into a temporary file, rather into memory. However, I was thinking that
> for removing 20bytes of information, reading and writing 15MB to 30MB is a
> lot of work. Is there a better way?

You can just read/write arbitrary-sized chunks in a loop, e.g.

	for (;;)
	{
		int n;
		lseek(fd, src_offset, SEEK_SET);
		n = read(fd, buf, sizeof(buf));
		if (!n)
			break;
		lseek(fd, dst_offset, SEEK_SET);
		write(fd, buf, n);
		src_offset += n;
		dst_offset += n;
	}
	ftruncate(fd, dst_offset);

Alternatively, you could mmap() the file and use memmove() instead of
the read/write pair (but not memcpy(); that isn't guaranteed to work
if source and destination overlap). That should provide better
performance; however, you shouldn't assume that you can mmap() the
entire file in one go, nor should you assume that both source and
destination will fit within a single mmap()'d region, so you would
need a pair of "scrolling windows". You would also have to deal with
the page-alignment issues.

BTW, if you were inserting (and hence shifting toward the end of the
file), you would have to start at the end of the file then work
backward.

-- 
Glynn Clements <glynn.clements@virgin.net>
-
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: quotactl

[Glynn Clements]

Darío Mariani wrote:

>    If you did not recompile the kernel, it has quota support. Has the FS 
> you are quering been mounted with the usrquota option?

The mount(8) manpage says (in regard of ext2):

       grpquota / noquota / quota / usrquota
              These options are accepted but ignored.

Examination of the kernel source code (fs/ext2/super.c) confirms this:

		/* Silently ignore the quota options */
		else if (!strcmp (this_char, "grpquota")
		         || !strcmp (this_char, "noquota")
		         || !strcmp (this_char, "quota")
		         || !strcmp (this_char, "usrquota"))
			/* Don't do anything ;-) */ ;

Also, if quota support is disabled (i.e. "!defined(CONFIG_QUOTA)"),
quotactl() will fail with ENOSYS rather than EINVAL (at least in
2.4.x).

-- 
Glynn Clements <glynn.clements@virgin.net>
-
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: quotactl

[Francis Lau]

> 
> Darío Mariani wrote:
> 
> >    If you did not recompile the kernel, it has quota support. Has the FS 
> > you are quering been mounted with the usrquota option?
> 
> The mount(8) manpage says (in regard of ext2):
> 
>        grpquota / noquota / quota / usrquota
>               These options are accepted but ignored.
> 
> Examination of the kernel source code (fs/ext2/super.c) confirms this:
> 
> 		/* Silently ignore the quota options */
> 		else if (!strcmp (this_char, "grpquota")
> 		         || !strcmp (this_char, "noquota")
> 		         || !strcmp (this_char, "quota")
> 		         || !strcmp (this_char, "usrquota"))
> 			/* Don't do anything ;-) */ ;
> 
> Also, if quota support is disabled (i.e. "!defined(CONFIG_QUOTA)"),
> quotactl() will fail with ENOSYS rather than EINVAL (at least in
> 2.4.x).
> 
> 
I have tried everything you guys suggested but am still unable to get 
quotactl to work.  For some reason, quotactl always returns a 0.  I have 
traced through fs/quota.c and I think the code always gets to the 'return 0' part 
in the do_quotactl function:

case Q_GETQUOTA: {
             struct if_dqblk idq;

            if ((ret = sb->s_qcop->get_dqblk(sb, type, id, &idq)))
                      return ret;
            if (copy_to_user(addr, &idq, sizeof(idq)))
                      return -EFAULT;
            return 0;
}

-----------------------------------
This is the kernel I am using:

[root@setup5 fs]# uname -a
Linux setup5 2.4.20-18.9smp #1 SMP Thu May 29 06:55:05 EDT 2003 i686 i686 
i386 GNU/Linux

----------------------------------
This is the test code I am using:

#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <linux/quota.h>

main ()
{
	int i, cmd;
	struct mem_dqblk mydq;

	cmd = QCMD(Q_GETQUOTA,USRQUOTA);
	i=quotactl (cmd, "/dev/md5", 70557, &mydq);

	printf ("curspace is %i\n", mydq.dqb_curspace);
}

----------------------------------
I made sure that CONFIG_QUOTA=y too.

[root@setup5 configs]# pwd
/usr/src/linux-2.4/configs
[root@setup5 configs]# grep CONFIG_QUOTA *
kernel-2.4.20-athlon.config:CONFIG_QUOTA=y
kernel-2.4.20-athlon-smp.config:CONFIG_QUOTA=y
kernel-2.4.20-i386-BOOT.config:# CONFIG_QUOTA is not set
kernel-2.4.20-i386-BOOT.config:# CONFIG_QUOTA is not set
kernel-2.4.20-i386.config:CONFIG_QUOTA=y
kernel-2.4.20-i386-smp.config:CONFIG_QUOTA=y
kernel-2.4.20-i586.config:CONFIG_QUOTA=y
kernel-2.4.20-i586-smp.config:CONFIG_QUOTA=y
kernel-2.4.20-i686-bigmem.config:CONFIG_QUOTA=y
kernel-2.4.20-i686.config:CONFIG_QUOTA=y
kernel-2.4.20-i686-smp.config:CONFIG_QUOTA=y
kernel-2.4.20-x86_64.config:CONFIG_QUOTA=y
kernel-2.4.20-x86_64-smp.config:CONFIG_QUOTA=y

---------------------------------
I ran a strace on the binary and this is what I got:

[root@setup5 C]# strace -f ./test
execve("./test", ["./test"], [/* 21 vars */]) = 0
uname({sys="Linux", node="setup5", ...}) = 0
brk(0)                                  = 0x8049628
old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 
0) = 0x40016000
open("/etc/ld.so.preload", O_RDONLY)    = -1 ENOENT (No such file or 
directory)
open("/etc/ld.so.cache", O_RDONLY)      = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=59724, ...}) = 0
old_mmap(NULL, 59724, PROT_READ, MAP_PRIVATE, 3, 0) = 0x40017000
close(3)                                = 0
open("/lib/tls/libc.so.6", O_RDONLY)    = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0`V\1B4\0"..., 512)
= 512
fstat64(3, {st_mode=S_IFREG|0755, st_size=1531064, ...}) = 0
old_mmap(0x42000000, 1257224, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 
0x42000000
old_mmap(0x4212e000, 12288, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 
3, 0x12e000) = 0x4212e000
old_mmap(0x42131000, 7944, PROT_READ|PROT_WRITE, 
MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x42131000
close(3)                                = 0
set_thread_area({entry_number:-1 -> 6, base_addr:0x400169e0, 
limit:1048575, seg_32bit:1, contents:0, read_exec_only:0, 
limit_in_pages:1, seg_not_present:0, useable:1}) = 0
munmap(0x40017000, 59724)               = 0
quotactl(0xd00 /* Q_??? */|USRQUOTA, "/dev/md5", 70557, {3221224176, 
1073790467, 1073830868, 1073832832, 1, 0, 1107383425, 134513434}) = -1 
EINVAL (Invalid argument)
fstat64(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 3), ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) 
= 0x40017000
write(1, "cmd is 851968\n", 14cmd is 851968
)         = 14
write(1, "i is -1\n", 8i is -1
)                = 8
write(1, "errno is 22\n", 12errno is 22
)           = 12
write(1, "curspace is 0\n", 14curspace is 0
)         = 14
munmap(0x40017000, 4096)                = 0
exit_group(14)



Thanks again.  I really appreciate it.

Francis

-
To unsubscribe from this list: send the line "unsubscribe linux-c-programming" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: quotactl

[Glynn Clements]

Francis Lau wrote:

> I have tried everything you guys suggested but am still unable to get 
> quotactl to work.  For some reason, quotactl always returns a 0.  I have 
> traced through fs/quota.c

fs/quota.c?

> and I think the code always gets to the 'return 0' part 
> in the do_quotactl function:

do_quotactl?

> case Q_GETQUOTA: {
>              struct if_dqblk idq;
> 
>             if ((ret = sb->s_qcop->get_dqblk(sb, type, id, &idq)))
>                       return ret;
>             if (copy_to_user(addr, &idq, sizeof(idq)))
>                       return -EFAULT;
>             return 0;
> }

I don't recognise this code, but looking around on the 'net suggests
that fs/quota.c and do_quotactl are from 2.5.x. I'm not using 2.5.x,
and I'm not about to download a 30Mb file to look into this.

Is this a vendor-modified kernel? Maybe the vendor has back-ported the
2.5.x quota code into their 2.4.x kernels for some reason (e.g. the
16-bit UID issue below)?

If you're using vendor-modified code, you might need to ask on a list
which is specific to that vendor's product. At the very least, you
need to give people a clue about which code you are using.

> This is the test code I am using:
> 
> #include <stdio.h>
> #include <stdlib.h>
> #include <errno.h>
> #include <linux/quota.h>
> 
> main ()
> {
> 	int i, cmd;
> 	struct mem_dqblk mydq;
> 
> 	cmd = QCMD(Q_GETQUOTA,USRQUOTA);
> 	i=quotactl (cmd, "/dev/md5", 70557, &mydq);

This won't work with the standard 2.4.x quota code; the UID has to fit
into 16 bits, otherwise the call will fail with EINVAL.

-- 
Glynn Clements <glynn.clements@virgin.net>