[linux-lvm] pvmove

[linux-lvm] pvmove

[Martin Papik]

Hi everyone

I'm new to the mailing list but not to LVM, until now everything was 
crystal clear and working fine. Now I'm either stuck or found a minor 
issue. I'm trying to move move extents on a physical volume. Not from 
one physical volume to another.

root@myhost# pvmove  /dev/md127:151808-152063 /dev/md127:152064-152319
   No extents available for allocation
root@myhost#

I also tried pvmove /dev/md127:151808-152063 /dev/md127:152064 and a few 
other combination. There are enough free extents. The volume I've 
created is 1G. There's 800 GB free. I made sure the destination blocks 
are available, I've tried different ranges.  Way below is the output of 
vgdisplay and a list of volumes and their extents.

Is this functionality (extent shuffling) supported? I mean is this a 
feature, bug or ignorance on my part?

PS, if you're wondering why would anyone want to move extents from a 
device to the very same device, you may notice in the output below that 
I have a few volumes spilt into pieces and I want to de-fragment it. I 
know it won't affect performance too much, probably at all. But it is 
going to help when (if?) I'm in dire need of dd if=/dev/xxx from the 
physical volume. Wouldn't be the first time (not the fault of LVM, 
disk+heat=bad blocks).

Thanks in advance for any ideas or scolding :-)

Martin

root@myhost# vgdisplay
   --- Volume group ---
   VG Name               volume
   System ID
   Format                lvm2
   Metadata Areas        1
   Metadata Sequence No  147
   VG Access             read/write
   VG Status             resizable
   MAX LV                0
   Cur LV                12
   Open LV               9
   Max PV                0
   Cur PV                1
   Act PV                1
   VG Size               3.35 TiB
   PE Size               4.00 MiB
   Total PE              878905
   Alloc PE / Size       666880 / 2.54 TiB
   Free  PE / Size       212025 / 828.22 GiB
   VG UUID Gv2rC8-sMgQ-xGGK-5swv-9Gck-QA1G-k50ABy

root@myhost# vgs --all -o 
+lv_name,lv_size,snap_percent,seg_pe_ranges,segtype,seg_size
   VG     #PV #LV #SN Attr   VSize VFree   LV      LSize Snap%  PE 
Ranges                Type   SSize
   volume   1  12   0 wz--n- 3.35t 828.22g home 150.00g        
/dev/md127:0-25599       linear 100.00g
   volume   1  12   0 wz--n- 3.35t 828.22g home 150.00g        
/dev/md127:72085-74495   linear   9.42g
   volume   1  12   0 wz--n- 3.35t 828.22g home 150.00g        
/dev/md127:309248-319636 linear  40.58g
   volume   1  12   0 wz--n- 3.35t 828.22g var 20.00g        
/dev/md127:25600-30719   linear  20.00g
   volume   1  12   0 wz--n- 3.35t 828.22g usr 20.00g        
/dev/md127:30720-35839   linear  20.00g
   volume   1  12   0 wz--n- 3.35t 828.22g tmp 2.00g        
/dev/md127:35840-36351   linear   2.00g
   volume   1  12   0 wz--n- 3.35t 828.22g p2p 300.00g        
/dev/md127:74496-151295  linear 300.00g
   volume   1  12   0 wz--n- 3.35t 828.22g boot 1.00g        
/dev/md127:151296-151551 linear   1.00g
   volume   1  12   0 wz--n- 3.35t 828.22g root 4.00g        
/dev/md127:152320-153343 linear   4.00g
   volume   1  12   0 wz--n- 3.35t 828.22g copy 2.00t        
/dev/md127:181248-309247 linear 500.00g
   volume   1  12   0 wz--n- 3.35t 828.22g copy 2.00t        
/dev/md127:606208-712298 linear 414.42g
   volume   1  12   0 wz--n- 3.35t 828.22g copy 2.00t        
/dev/md127:156672-181247 linear  96.00g
   volume   1  12   0 wz--n- 3.35t 828.22g copy 2.00t        
/dev/md127:36352-72084   linear 139.58g
   volume   1  12   0 wz--n- 3.35t 828.22g copy 2.00t        
/dev/md127:319637-549524 linear 898.00g
   volume   1  12   0 wz--n- 3.35t 828.22g swap 8.00g        
/dev/md127:153344-155391 linear   8.00g
   volume   1  12   0 wz--n- 3.35t 828.22g emulate 50.00g        
/dev/md127:549525-562324 linear  50.00g
   volume   1  12   0 wz--n- 3.35t 828.22g test 1.00g        
/dev/md127:151552-151807 linear   1.00g
   volume   1  12   0 wz--n- 3.35t 828.22g test2 1.00g        
/dev/md127:151808-152063 linear   1.00g
root@myhost#

Re: [linux-lvm] pvmove

[Zdenek Kabelac]

Dne 19.7.2013 20:14, Martin Papik napsal(a):
> Hi everyone
>
> I'm new to the mailing list but not to LVM, until now everything was crystal
> clear and working fine. Now I'm either stuck or found a minor issue. I'm
> trying to move move extents on a physical volume. Not from one physical volume
> to another.
>
> root@myhost# pvmove  /dev/md127:151808-152063 /dev/md127:152064-152319
>    No extents available for allocation
> root@myhost#

Add  --alloc anywhere

But I assume the tool could be a bit more smarter here.
It looks more or less like a bug to me.

Zdenek

Re: [linux-lvm] pvmove

[Martin Papik]

Dear Zdenek

Thank you very much, it works. I noticed a few more things. Not errors, 
just minor things that might be documented better or handled 
differently. Depending on the design philosophy.

"pvmove /dev/md127:152064-152319 /dev/md127:800000 --alloc anywhere" 
fails because /dev/md127:xx seems to be interpreted as "one extent 
starting at xx", I'm not sure that's what I'd guess from the man page.

"pvmove /dev/md127:152064-152319 /dev/md127:800000- --alloc anywhere" 
works, but the manual page indicates that the syntax is 
DestinationPhysicalVolume[:PE[-PE]...], which I would interpret as 
/dev/xx:111 OR /dev/xx:111-222, but not /dev/xx:111-, that would IMHO be 
DestinationPhysicalVolume[:PE[-[PE]]...].

I'd interpret /dev/md126:100 as a start of an area, since it has no 
explicit bound, where as /dev/md127:100-100 I would expect for a single 
extent, since there is an explicit bound. Does that make sense?

PS, how about a syntax variant where it can be /dev/md127:100+30 meaning 
30 extents starting at 100? I understand it's often better to print 
ranges, but maybe in some cases start+length might work better. Or 
start-end+length for printing, and start[-end][+length] as input, and 
exit with error if both end and length are specified and they don't 
match (start+length!=length).

I'm not complaining, just suggesting enhancements. :-)

Again, thank you very much for your quick and helpful response.

Martin

On 07/19/2013 10:59 PM, Zdenek Kabelac wrote:
> Dne 19.7.2013 20:14, Martin Papik napsal(a):
>> Hi everyone
>>
>> I'm new to the mailing list but not to LVM, until now everything was 
>> crystal
>> clear and working fine. Now I'm either stuck or found a minor issue. I'm
>> trying to move move extents on a physical volume. Not from one 
>> physical volume
>> to another.
>>
>> root@myhost# pvmove  /dev/md127:151808-152063 /dev/md127:152064-152319
>>    No extents available for allocation
>> root@myhost#
>
> Add  --alloc anywhere
>
> But I assume the tool could be a bit more smarter here.
> It looks more or less like a bug to me.
>
> Zdenek
>

Re: [linux-lvm] pvmove

[Alasdair G Kergon]

It uses lists of ranges separated by colons.

  :x-y

If x is missing, 0 is inserted; if y is missing, the last extent on the device
is inserted.

Man pages can always be improved with more examples!

For +length, try the patch below.

Alasdair


--- a/tools/toollib.c
+++ b/tools/toollib.c
@@ -1001,7 +1001,7 @@ static int _parse_pes(struct dm_pool *mem, char *c, struct dm_list *pe_ranges,
 		      const char *pvname, uint32_t size)
 {
 	char *endptr;
-	uint32_t start, end;
+	uint32_t start, end, len;
 
 	/* Default to whole PV */
 	if (!c) {
@@ -1041,7 +1041,16 @@ static int _parse_pes(struct dm_pool *mem, char *c, struct dm_list *pe_ranges,
 					goto error;
 				c = endptr;
 			}
+		} else if (*c == '+') {	/* Length? */
+			c++;
+			if (isdigit(*c)) {
+				if (!xstrtouint32(c, &endptr, 10, &len))
+					goto error;
+				c = endptr;
+				end = start + (len ? (len - 1) : 0);
+			}
 		}
+
 		if (*c && *c != ':')
 			goto error;
 

Re: [linux-lvm] pvmove

[Martin Papik]

[-- Attachment #1: Type: text/plain, Size: 8000 bytes --]


Thank you. Is that patch going to be added to the main source code base?

I think I found an error (segfault), see below.

==> I have a few more improvement suggestions.

* IMHO vgs/lvs should include a column called seg_size_pe, similar to 
seg_start and seg_start_pe
* IMHO vgs/lvs/pvs should have an command line parameter, or option 
modifier allowing to not display empty columns, i.e. the following 
command might skip displaying origin, snap%, move, log, copy%, convert, 
because they have no values. That would allow me to show copy_percent 
column only when copy is in progress. Would be nice.

lvs --all -o +vg_extent_size
   LV      VG     Attr   LSize   Origin Snap%  Move Log Copy%  Convert Ext
   boot    volume -wi-ao 1.00g                                       4.00m
   copy    volume -wi-ao 2.00t                                       4.00m
   emulate volume -wi-ao 50.00g                                       4.00m
   home    volume -wi-ao 150.00g                                       4.00m
   p2p     volume -wi-a- 300.00g                                       4.00m
   root    volume -wi-ao 4.00g                                       4.00m
   swap    volume -wi-ao 8.00g                                       4.00m
   tmp     volume -wi-ao 2.00g                                       4.00m
   usr     volume -wi-ao 20.00g                                       4.00m
   var     volume -wi-ao 20.00g                                       4.00m

Possible error:

root@localhost# lvs --rows --align
Does not align, should it?

*==> SEGFAULT INFO*

root@localhost# pvs --all -o +vg_fmt
Segmentation fault (core dumped)

However the following seems to work.....

root@localhost# pvs --all -o +vg_name
   PV         VG     Fmt  Attr PSize PFree VG
   /dev/dm-0              --      0     0
   --- some entries removed for brevity ----
   /dev/md127 volume lvm2 a-   5.46t 2.77t volume
   --- some entries removed for brevity ----
   /dev/sde1              --      0     0

Additional information that might help.

root@localhost# pvs --version
   LVM version:     2.02.66(2) (2010-05-20)
   Library version: 1.02.48 (2010-05-20)
   Driver version:  4.22.0
root@localhost# cat /etc/lsb-release
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=12.04
DISTRIB_CODENAME=precise
DISTRIB_DESCRIPTION="Ubuntu 12.04.2 LTS"

root@localhost#valgrind pvs --all -o +vg_fmt
==14584== Memcheck, a memory error detector
==14584== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
==14584== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
==14584== Command: pvs --all -o +vg_fmt
==14584==
==14584== Invalid read of size 8
==14584==    at 0x548ED77: dm_report_field_string (in 
/lib/libdevmapper.so.1.02.1)
==14584==    by 0x549006F: dm_report_object (in /lib/libdevmapper.so.1.02.1)
==14584==    by 0x46B138: report_object (in /sbin/lvm)
==14584==    by 0x423618: ??? (in /sbin/lvm)
==14584==    by 0x4254B2: process_each_pv (in /sbin/lvm)
==14584==    by 0x4230D3: ??? (in /sbin/lvm)
==14584==    by 0x41AE60: lvm_run_command (in /sbin/lvm)
==14584==    by 0x41B307: lvm2_main (in /sbin/lvm)
==14584==    by 0x58CE76C: (below main) (libc-start.c:226)
==14584==  Address 0x28 is not stack'd, malloc'd or (recently) free'd
==14584==
==14584==
==14584== Process terminating with default action of signal 11 (SIGSEGV)
==14584==  Access not within mapped region at address 0x28
==14584==    at 0x548ED77: dm_report_field_string (in 
/lib/libdevmapper.so.1.02.1)
==14584==    by 0x549006F: dm_report_object (in /lib/libdevmapper.so.1.02.1)
==14584==    by 0x46B138: report_object (in /sbin/lvm)
==14584==    by 0x423618: ??? (in /sbin/lvm)
==14584==    by 0x4254B2: process_each_pv (in /sbin/lvm)
==14584==    by 0x4230D3: ??? (in /sbin/lvm)
==14584==    by 0x41AE60: lvm_run_command (in /sbin/lvm)
==14584==    by 0x41B307: lvm2_main (in /sbin/lvm)
==14584==    by 0x58CE76C: (below main) (libc-start.c:226)
==14584==  If you believe this happened as a result of a stack
==14584==  overflow in your program's main thread (unlikely but
==14584==  possible), you can try to increase the size of the
==14584==  main thread stack using the --main-stacksize= flag.
==14584==  The main thread stack size used in this run was 8388608.
==14584==

root@localhost# ltrace pvs --all -o +vg_fmt 2>&1 | tail
__ctype_toupper_loc()                            = 0x7f2f0e83f768
__ctype_toupper_loc()                            = 0x7f2f0e83f768
__ctype_toupper_loc()                            = 0x7f2f0e83f768
__sprintf_chk(0x1e86e88, 1, -1, 0x48daa5, 0x48dab8) = 2
dm_pool_strdup(0x1e88120, 0x1e86e88, 0x1e86e8a, 0, 0x48daa8) = 0x1e8e538
dm_pool_alloc(0x1e88120, 8, 2, 0x7f2f0d892030, 0x48daa8) = 0x1e8e540
dm_report_field_set_value(0x1e8e508, 0x1e8e538, 0x1e8e540, 
0x7f2f0d892030, 0x48daa8) = 0x1e8e540
dm_report_field_string(0x1e88080, 0x1e8e550, 40, 0x6b1fe0, 0x1e6cfe0 
<unfinished ...>
--- SIGSEGV (Segmentation fault) ---
+++ killed by SIGSEGV +++
root@localhost#

root@localhost# gdb --args pvs --all -o +vg_fmt
GNU gdb (Ubuntu/Linaro 7.4-2012.04-0ubuntu2.1) 7.4-2012.04
Copyright (C) 2012 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later 
<http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
For bug reporting instructions, please see:
<http://bugs.launchpad.net/gdb-linaro/>...
Reading symbols from /sbin/pvs...(no debugging symbols found)...done.
(gdb) run
Starting program: /sbin/pvs --all -o +vg_fmt
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
File descriptor 3 (socket:[46992174]) leaked on pvs invocation. Parent 
PID 14623: gdb
File descriptor 4 (socket:[46992175]) leaked on pvs invocation. Parent 
PID 14623: gdb
File descriptor 5 (pipe:[46992176]) leaked on pvs invocation. Parent PID 
14623: gdb
File descriptor 6 (pipe:[46992176]) leaked on pvs invocation. Parent PID 
14623: gdb

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff757dd77 in dm_report_field_string () from 
/lib/libdevmapper.so.1.02.1
(gdb) disassemble dm_report_field_string
Dump of assembler code for function dm_report_field_string:
    0x00007ffff757dd70 <+0>:    push   %rbx
    0x00007ffff757dd71 <+1>:    mov    %rsi,%rbx
    0x00007ffff757dd74 <+4>:    mov    (%rdi),%rdi
=> 0x00007ffff757dd77 <+7>:    mov (%rdx),%rsi
    0x00007ffff757dd7a <+10>:    callq 0x7ffff7570a30 <dm_pool_strdup@plt>
    0x00007ffff757dd7f <+15>:    test   %rax,%rax
(gdb) info registers
rax            0x6dd550    7198032
rbx            0x6dd550    7198032
rcx            0x6b1fe0    7020512
rdx            0x28    40
rsi            0x6dd550    7198032
rdi            0x6d7120    7172384
rbp            0x6d7080    0x6d7080



On 07/20/2013 02:20 AM, Alasdair G Kergon wrote:
> It uses lists of ranges separated by colons.
>
>    :x-y
>
> If x is missing, 0 is inserted; if y is missing, the last extent on the device
> is inserted.
>
> Man pages can always be improved with more examples!
>
> For +length, try the patch below.
>
> Alasdair
>
>
> --- a/tools/toollib.c
> +++ b/tools/toollib.c
> @@ -1001,7 +1001,7 @@ static int _parse_pes(struct dm_pool *mem, char *c, struct dm_list *pe_ranges,
>   		      const char *pvname, uint32_t size)
>   {
>   	char *endptr;
> -	uint32_t start, end;
> +	uint32_t start, end, len;
>   
>   	/* Default to whole PV */
>   	if (!c) {
> @@ -1041,7 +1041,16 @@ static int _parse_pes(struct dm_pool *mem, char *c, struct dm_list *pe_ranges,
>   					goto error;
>   				c = endptr;
>   			}
> +		} else if (*c == '+') {	/* Length? */
> +			c++;
> +			if (isdigit(*c)) {
> +				if (!xstrtouint32(c, &endptr, 10, &len))
> +					goto error;
> +				c = endptr;
> +				end = start + (len ? (len - 1) : 0);
> +			}
>   		}
> +
>   		if (*c && *c != ':')
>   			goto error;
>   


[-- Attachment #2: Type: text/html, Size: 16750 bytes --]

Re: [linux-lvm] pvmove ==> segfault: pvs --all -o +vg_fmt

[Martin Papik]

[-- Attachment #1: Type: text/plain, Size: 8465 bytes --]


No reply yet so I thought I'd modify the subject to catch someone's eye.

Martin

On 07/23/2013 05:45 AM, Martin Papik wrote:
>
> Thank you. Is that patch going to be added to the main source code base?
>
> I think I found an error (segfault), see below.
>
> ==> I have a few more improvement suggestions.
>
> * IMHO vgs/lvs should include a column called seg_size_pe, similar to 
> seg_start and seg_start_pe
> * IMHO vgs/lvs/pvs should have an command line parameter, or option 
> modifier allowing to not display empty columns, i.e. the following 
> command might skip displaying origin, snap%, move, log, copy%, 
> convert, because they have no values. That would allow me to show 
> copy_percent column only when copy is in progress. Would be nice.
>
> lvs --all -o +vg_extent_size
>   LV      VG     Attr   LSize   Origin Snap%  Move Log Copy%  Convert Ext
>   boot    volume -wi-ao 1.00g                                       4.00m
>   copy    volume -wi-ao 2.00t                                       4.00m
>   emulate volume -wi-ao 50.00g                                       4.00m
>   home    volume -wi-ao 150.00g                                       
> 4.00m
>   p2p     volume -wi-a- 300.00g                                       
> 4.00m
>   root    volume -wi-ao 4.00g                                       4.00m
>   swap    volume -wi-ao 8.00g                                       4.00m
>   tmp     volume -wi-ao 2.00g                                       4.00m
>   usr     volume -wi-ao 20.00g                                       4.00m
>   var     volume -wi-ao 20.00g                                       4.00m
>
> Possible error:
>
> root@localhost# lvs --rows --align
> Does not align, should it?
>
> *==> SEGFAULT INFO*
>
> root@localhost# pvs --all -o +vg_fmt
> Segmentation fault (core dumped)
>
> However the following seems to work.....
>
> root@localhost# pvs --all -o +vg_name
>   PV         VG     Fmt  Attr PSize PFree VG
>   /dev/dm-0              --      0     0
>   --- some entries removed for brevity ----
>   /dev/md127 volume lvm2 a-   5.46t 2.77t volume
>   --- some entries removed for brevity ----
>   /dev/sde1              --      0     0
>
> Additional information that might help.
>
> root@localhost# pvs --version
>   LVM version:     2.02.66(2) (2010-05-20)
>   Library version: 1.02.48 (2010-05-20)
>   Driver version:  4.22.0
> root@localhost# cat /etc/lsb-release
> DISTRIB_ID=Ubuntu
> DISTRIB_RELEASE=12.04
> DISTRIB_CODENAME=precise
> DISTRIB_DESCRIPTION="Ubuntu 12.04.2 LTS"
>
> root@localhost#valgrind pvs --all -o +vg_fmt
> ==14584== Memcheck, a memory error detector
> ==14584== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
> ==14584== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright 
> info
> ==14584== Command: pvs --all -o +vg_fmt
> ==14584==
> ==14584== Invalid read of size 8
> ==14584==    at 0x548ED77: dm_report_field_string (in 
> /lib/libdevmapper.so.1.02.1)
> ==14584==    by 0x549006F: dm_report_object (in 
> /lib/libdevmapper.so.1.02.1)
> ==14584==    by 0x46B138: report_object (in /sbin/lvm)
> ==14584==    by 0x423618: ??? (in /sbin/lvm)
> ==14584==    by 0x4254B2: process_each_pv (in /sbin/lvm)
> ==14584==    by 0x4230D3: ??? (in /sbin/lvm)
> ==14584==    by 0x41AE60: lvm_run_command (in /sbin/lvm)
> ==14584==    by 0x41B307: lvm2_main (in /sbin/lvm)
> ==14584==    by 0x58CE76C: (below main) (libc-start.c:226)
> ==14584==  Address 0x28 is not stack'd, malloc'd or (recently) free'd
> ==14584==
> ==14584==
> ==14584== Process terminating with default action of signal 11 (SIGSEGV)
> ==14584==  Access not within mapped region at address 0x28
> ==14584==    at 0x548ED77: dm_report_field_string (in 
> /lib/libdevmapper.so.1.02.1)
> ==14584==    by 0x549006F: dm_report_object (in 
> /lib/libdevmapper.so.1.02.1)
> ==14584==    by 0x46B138: report_object (in /sbin/lvm)
> ==14584==    by 0x423618: ??? (in /sbin/lvm)
> ==14584==    by 0x4254B2: process_each_pv (in /sbin/lvm)
> ==14584==    by 0x4230D3: ??? (in /sbin/lvm)
> ==14584==    by 0x41AE60: lvm_run_command (in /sbin/lvm)
> ==14584==    by 0x41B307: lvm2_main (in /sbin/lvm)
> ==14584==    by 0x58CE76C: (below main) (libc-start.c:226)
> ==14584==  If you believe this happened as a result of a stack
> ==14584==  overflow in your program's main thread (unlikely but
> ==14584==  possible), you can try to increase the size of the
> ==14584==  main thread stack using the --main-stacksize= flag.
> ==14584==  The main thread stack size used in this run was 8388608.
> ==14584==
>
> root@localhost# ltrace pvs --all -o +vg_fmt 2>&1 | tail
> __ctype_toupper_loc()                            = 0x7f2f0e83f768
> __ctype_toupper_loc()                            = 0x7f2f0e83f768
> __ctype_toupper_loc()                            = 0x7f2f0e83f768
> __sprintf_chk(0x1e86e88, 1, -1, 0x48daa5, 0x48dab8) = 2
> dm_pool_strdup(0x1e88120, 0x1e86e88, 0x1e86e8a, 0, 0x48daa8) = 0x1e8e538
> dm_pool_alloc(0x1e88120, 8, 2, 0x7f2f0d892030, 0x48daa8) = 0x1e8e540
> dm_report_field_set_value(0x1e8e508, 0x1e8e538, 0x1e8e540, 
> 0x7f2f0d892030, 0x48daa8) = 0x1e8e540
> dm_report_field_string(0x1e88080, 0x1e8e550, 40, 0x6b1fe0, 0x1e6cfe0 
> <unfinished ...>
> --- SIGSEGV (Segmentation fault) ---
> +++ killed by SIGSEGV +++
> root@localhost#
>
> root@localhost# gdb --args pvs --all -o +vg_fmt
> GNU gdb (Ubuntu/Linaro 7.4-2012.04-0ubuntu2.1) 7.4-2012.04
> Copyright (C) 2012 Free Software Foundation, Inc.
> License GPLv3+: GNU GPL version 3 or later 
> <http://gnu.org/licenses/gpl.html>
> This is free software: you are free to change and redistribute it.
> There is NO WARRANTY, to the extent permitted by law. Type "show copying"
> and "show warranty" for details.
> This GDB was configured as "x86_64-linux-gnu".
> For bug reporting instructions, please see:
> <http://bugs.launchpad.net/gdb-linaro/>...
> Reading symbols from /sbin/pvs...(no debugging symbols found)...done.
> (gdb) run
> Starting program: /sbin/pvs --all -o +vg_fmt
> [Thread debugging using libthread_db enabled]
> Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
> File descriptor 3 (socket:[46992174]) leaked on pvs invocation. Parent 
> PID 14623: gdb
> File descriptor 4 (socket:[46992175]) leaked on pvs invocation. Parent 
> PID 14623: gdb
> File descriptor 5 (pipe:[46992176]) leaked on pvs invocation. Parent 
> PID 14623: gdb
> File descriptor 6 (pipe:[46992176]) leaked on pvs invocation. Parent 
> PID 14623: gdb
>
> Program received signal SIGSEGV, Segmentation fault.
> 0x00007ffff757dd77 in dm_report_field_string () from 
> /lib/libdevmapper.so.1.02.1
> (gdb) disassemble dm_report_field_string
> Dump of assembler code for function dm_report_field_string:
>    0x00007ffff757dd70 <+0>:    push   %rbx
>    0x00007ffff757dd71 <+1>:    mov    %rsi,%rbx
>    0x00007ffff757dd74 <+4>:    mov    (%rdi),%rdi
> => 0x00007ffff757dd77 <+7>:    mov (%rdx),%rsi
>    0x00007ffff757dd7a <+10>:    callq 0x7ffff7570a30 <dm_pool_strdup@plt>
>    0x00007ffff757dd7f <+15>:    test   %rax,%rax
> (gdb) info registers
> rax            0x6dd550    7198032
> rbx            0x6dd550    7198032
> rcx            0x6b1fe0    7020512
> rdx            0x28    40
> rsi            0x6dd550    7198032
> rdi            0x6d7120    7172384
> rbp            0x6d7080    0x6d7080
>
>
>
> On 07/20/2013 02:20 AM, Alasdair G Kergon wrote:
>> It uses lists of ranges separated by colons.
>>
>>    :x-y
>>
>> If x is missing, 0 is inserted; if y is missing, the last extent on the device
>> is inserted.
>>
>> Man pages can always be improved with more examples!
>>
>> For +length, try the patch below.
>>
>> Alasdair
>>
>>
>> --- a/tools/toollib.c
>> +++ b/tools/toollib.c
>> @@ -1001,7 +1001,7 @@ static int _parse_pes(struct dm_pool *mem, char *c, struct dm_list *pe_ranges,
>>   		      const char *pvname, uint32_t size)
>>   {
>>   	char *endptr;
>> -	uint32_t start, end;
>> +	uint32_t start, end, len;
>>   
>>   	/* Default to whole PV */
>>   	if (!c) {
>> @@ -1041,7 +1041,16 @@ static int _parse_pes(struct dm_pool *mem, char *c, struct dm_list *pe_ranges,
>>   					goto error;
>>   				c = endptr;
>>   			}
>> +		} else if (*c == '+') {	/* Length? */
>> +			c++;
>> +			if (isdigit(*c)) {
>> +				if (!xstrtouint32(c, &endptr, 10, &len))
>> +					goto error;
>> +				c = endptr;
>> +				end = start + (len ? (len - 1) : 0);
>> +			}
>>   		}
>> +
>>   		if (*c && *c != ':')
>>   			goto error;
>>   
>


[-- Attachment #2: Type: text/html, Size: 17693 bytes --]

Re: [linux-lvm] pvmove

[Alasdair G Kergon]

On Tue, Jul 23, 2013 at 05:45:50AM +0300, Martin Papik wrote:
> Thank you. Is that patch going to be added to the main source code base?

Committed for 2.02.103.

> I think I found an error (segfault), see below.

>   LVM version:     2.02.66(2) (2010-05-20)

I've no interest in bugs reported against such an old version of the code, I'm
afraid.  If you reproduce with upstream code, then we'll look into it.

Alasdair

Re: [linux-lvm] pvmove

[Alasdair G Kergon]

On Tue, Jul 23, 2013 at 05:45:50AM +0300, Martin Papik wrote:
> ==> I have a few more improvement suggestions.
> * IMHO vgs/lvs should include a column called seg_size_pe, similar to  
> seg_start and seg_start_pe

Implemented for 2.02.103.
  https://git.fedorahosted.org/cgit/lvm2.git/commit/?id=11dc6a03c4cdd10bc4065c7d7a8585373455f947

Alasdair

Re: [linux-lvm] pvmove

[Alasdair G Kergon]

On Tue, Jul 23, 2013 at 05:45:50AM +0300, Martin Papik wrote:
> * IMHO vgs/lvs/pvs should have an command line parameter, or option  
> modifier allowing to not display empty columns, i.e. the following  
> command might skip displaying origin, snap%, move, log, copy%, convert,  
> because they have no values. That would allow me to show copy_percent  
> column only when copy is in progress. Would be nice.

Yes - I'm surprised we never had this suggested before.

- Some columns will gain a property saying they can be suppressed when empty.

- A new cmdline option/config setting to suppress empty fields.

- Implemented as an extension alongside the 'sort' option.  (Like sort, you
  have to have all the output data prepared before you find out whether any
  columns are empty - even the headings can't be output before this point.)

- A way to flag an output data entry as empty.  (zkabelac is about to commit
  some patches to tidy up the handling of empty fields so this change can
  follow those.)

Alasdair

[linux-lvm] pvmove

[Juan Pablo Giménez]

[-- Attachment #1.1: Type: text/plain, Size: 897 bytes --]

What's more secure, the Redhat patched pvmove or the original one?

BTW, I have a little patch to fix RedHat patched pvmove and make it work
with unpatched kernels...


        --- lib/pv_move.c	2003-07-11 19:10:26.000000000 -0300
        +++ lib/pv_move.c.rcom	2003-07-11 19:09:35.000000000 -0300
        @@ -916,7 +916,7 @@
         	ret = ioctl(group, PE_LOCKED_COPY, &pe_copy_req);
         	if (ret < 0)
         		ret = -errno;
        -	if (ret == -EINVAL)
        +	if (ret == -EINVAL || ret == -ENOTTY)
         		ret = -LVM_EPV_LOCKED_COPY_EINVAL;
         	
         	close(group);
        




Juan Pablo Giménez 

http://www.rcom.com.ar/
http://www.rcom.com.ar/comunidad/
Llave GnuPG:
MIT keyserver 
Web Rcom 

GnuPG fingerprint: ED97 6FB8 A9BE
26A6 789B 6657 0A15 EC2B D3D8 FD2B 

http://www.greenpeace.org.ar/secciones/cyberactivismo/nuclear.php3

[-- Attachment #1.2: Type: text/html, Size: 2027 bytes --]

[-- Attachment #2: Esta parte del mensaje está firmada digitalmente --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [linux-lvm] pvmove

[Joe Thornber]

On Fri, Jul 11, 2003 at 07:12:26PM -0300, Juan Pablo Gim?nez wrote:
> What's more secure, the Redhat patched pvmove or the original one?

The origininal pvmove has some serious problems that can make it
deadlock.  Redhats version attempts to correct these.

The most recent LVM2 tools (see Alasdairs recent email) contain a
pvmove implementation as well.

- Joe