[9fans] Corrupted file entry on QEMU - how to recover?

[9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

Hello. I got a strange cacheLocalAlloc function error consisting of  
"get type 0 exp 0...type 40509443? exp 0". I did an fshalt and  
restarted, but now I get

	user[none]: pietro
	time...
	fossil(#S/sdC0/fossil)...fsOpen error
	fsOpen: fileRoot: corrupted file entry
	fsys main open -V -c 3000: fsOpen: fileRoot: corrupted file entry
	version...boot: mount /: fsys 'main' not open
	panic: boot process died: unknown
	panic: boot process died: unknown
	dumpstack disabled
	cpu0: exiting

Note that the boot process died message comes twice. Every time I  
subsequently try to start Plan 9, I get this error. What's going on,  
and can I use the installer disk to fix it, and how?

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Skip Tavakkolian]

> Hello. I got a strange cacheLocalAlloc function error consisting of  
> "get type 0 exp 0...type 40509443? exp 0". I did an fshalt and  
> restarted, but now I get
> 
> 	user[none]: pietro
> 	time...
> 	fossil(#S/sdC0/fossil)...fsOpen error
> 	fsOpen: fileRoot: corrupted file entry
> 	fsys main open -V -c 3000: fsOpen: fileRoot: corrupted file entry
> 	version...boot: mount /: fsys 'main' not open
> 	panic: boot process died: unknown
> 	panic: boot process died: unknown
> 	dumpstack disabled
> 	cpu0: exiting
> 
> Note that the boot process died message comes twice. Every time I  
> subsequently try to start Plan 9, I get this error. What's going on,  
> and can I use the installer disk to fix it, and how?

you should be able to boot from an install cd and then
start another fossil for /dev/sdC0/fossil to see what's
going on.

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I'm downloading the latest install disc right now, but how do I get a  
fossil onto the system? mount -m/n/myfossl /dev/sdC0/fossil?

On Dec 16, 2007, at 4:10 AM, Skip Tavakkolian wrote:

>> Hello. I got a strange cacheLocalAlloc function error consisting of
>> "get type 0 exp 0...type 40509443? exp 0". I did an fshalt and
>> restarted, but now I get
>>
>> 	user[none]: pietro
>> 	time...
>> 	fossil(#S/sdC0/fossil)...fsOpen error
>> 	fsOpen: fileRoot: corrupted file entry
>> 	fsys main open -V -c 3000: fsOpen: fileRoot: corrupted file entry
>> 	version...boot: mount /: fsys 'main' not open
>> 	panic: boot process died: unknown
>> 	panic: boot process died: unknown
>> 	dumpstack disabled
>> 	cpu0: exiting
>>
>> Note that the boot process died message comes twice. Every time I
>> subsequently try to start Plan 9, I get this error. What's going on,
>> and can I use the installer disk to fix it, and how?
>
> you should be able to boot from an install cd and then
> start another fossil for /dev/sdC0/fossil to see what's
> going on.
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[erik quanstrom]

> I'm downloading the latest install disc right now, but how do I get a  
> fossil onto the system? mount -m/n/myfossl /dev/sdC0/fossil?

you likely want to use fossil/flcheck and fossil/conf to see if the
configuration looks sane first.

fossil typically reads its configuration when given the -f option.
you might not want that.  if things are goofy, it might be easier
to start up a fossil with no commands except one to start
a console.

	fossil -c  'srv -p rcons' 

and then enter configuration by hand and fossil's console.  
you can use fossil/conf to see what commands are normally used
to start fossil.

the only problem with this way of debugging things is you need
to remember that the kernel adds a few commands when it
boots:

	; g /boot/fossil /sys/src/9/boot/local.c|grep run
	/sys/src/9/boot/local.c:254: 	run("/boot/fossil", "-f", partition, "-c", "srv -A fboot", "-c", "srv -p fscons", 0);

if your fossil's really corrupt, perhaps you have a dma issue?

- erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

fossil/flchk expects a venti server and I don't know how to tell it  
to suppress one (-h something) since I have insufficient disk space  
for venti. As for rcons, every time I try the commands fossil/conf  
spits out, the last one always exits with the errors I described  
before and then "fsys main check" doesn't work ("fsys main not open").

On Dec 16, 2007, at 11:25 AM, erik quanstrom wrote:

>> I'm downloading the latest install disc right now, but how do I get a
>> fossil onto the system? mount -m/n/myfossl /dev/sdC0/fossil?
>
> you likely want to use fossil/flcheck and fossil/conf to see if the
> configuration looks sane first.
>
> fossil typically reads its configuration when given the -f option.
> you might not want that.  if things are goofy, it might be easier
> to start up a fossil with no commands except one to start
> a console.
>
> 	fossil -c  'srv -p rcons'
>
> and then enter configuration by hand and fossil's console.
> you can use fossil/conf to see what commands are normally used
> to start fossil.
>
> the only problem with this way of debugging things is you need
> to remember that the kernel adds a few commands when it
> boots:
>
> 	; g /boot/fossil /sys/src/9/boot/local.c|grep run
> 	/sys/src/9/boot/local.c:254: 	run("/boot/fossil", "-f", partition,  
> "-c", "srv -A fboot", "-c", "srv -p fscons", 0);
>
> if your fossil's really corrupt, perhaps you have a dma issue?
>
> - erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[erik quanstrom]

> fossil/flchk expects a venti server and I don't know how to tell it  
> to suppress one (-h something) since I have insufficient disk space  
> for venti. As for rcons, every time I try the commands fossil/conf  
> spits out, the last one always exits with the errors I described  
> before and then "fsys main check" doesn't work ("fsys main not open").

you likely need to check the outout of
	fossil/conf /dev/sdC0/fossil
it seems likely that venti is specified in your fossil configuration.

- erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

It actually says that venti should be ignored (-V). flchk still gives  
me the error, though. I tried  -h '' and it didn't work. -V wasn't  
supported either, and in rcons, fsys main check fix refused to work  
unless the system was open, which it wasn't!

On Dec 16, 2007, at 12:25 PM, erik quanstrom wrote:

>> fossil/flchk expects a venti server and I don't know how to tell it
>> to suppress one (-h something) since I have insufficient disk space
>> for venti. As for rcons, every time I try the commands fossil/conf
>> spits out, the last one always exits with the errors I described
>> before and then "fsys main check" doesn't work ("fsys main not  
>> open").
>
> you likely need to check the outout of
> 	fossil/conf /dev/sdC0/fossil
> it seems likely that venti is specified in your fossil configuration.
>
> - erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

Good news: I just found an -f option in flchk that turns off venti!  
Going to try.

On Dec 16, 2007, at 12:25 PM, erik quanstrom wrote:

>> fossil/flchk expects a venti server and I don't know how to tell it
>> to suppress one (-h something) since I have insufficient disk space
>> for venti. As for rcons, every time I try the commands fossil/conf
>> spits out, the last one always exits with the errors I described
>> before and then "fsys main check" doesn't work ("fsys main not  
>> open").
>
> you likely need to check the outout of
> 	fossil/conf /dev/sdC0/fossil
> it seems likely that venti is specified in your fossil configuration.
>
> - erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

And now it still says it can't open the file system. What's wrong  
with checking the raw data? I don't want to lose my files, and  
nothing seems to work without calling fsOpen.

On Dec 16, 2007, at 8:34 PM, Pietro Gagliardi wrote:

> Good news: I just found an -f option in flchk that turns off venti!  
> Going to try.
>
> On Dec 16, 2007, at 12:25 PM, erik quanstrom wrote:
>
>>> fossil/flchk expects a venti server and I don't know how to tell it
>>> to suppress one (-h something) since I have insufficient disk space
>>> for venti. As for rcons, every time I try the commands fossil/conf
>>> spits out, the last one always exits with the errors I described
>>> before and then "fsys main check" doesn't work ("fsys main not  
>>> open").
>>
>> you likely need to check the outout of
>> 	fossil/conf /dev/sdC0/fossil
>> it seems likely that venti is specified in your fossil configuration.
>>
>> - erik
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[erik quanstrom]

On Sun Dec 16 20:37:03 EST 2007, pietro10@mac.com wrote:
> And now it still says it can't open the file system. What's wrong  
> with checking the raw data? I don't want to lose my files, and  
> nothing seems to work without calling fsOpen.

fossil is just a write buffer.  if fossil thinks it's been written to
venti, then fossil would have no access without a venti server.

so it would seem to me that the simplist explination for this is
that although you think you haven't configured venti, somehow
it was.

do you have a venti paritition?  also the output of venti/conf
would be helpful.

- erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I have no venti partition, and venti/conf asks for one called  
v.arenas. The reason was that the default configuration of the  
installer only gave fossil 700MB and venti/arenas 3GB. I wasn't sure  
that would be sufficient for my files :-)

On Dec 16, 2007, at 8:42 PM, erik quanstrom wrote:

> On Sun Dec 16 20:37:03 EST 2007, pietro10@mac.com wrote:
>> And now it still says it can't open the file system. What's wrong
>> with checking the raw data? I don't want to lose my files, and
>> nothing seems to work without calling fsOpen.
>
> fossil is just a write buffer.  if fossil thinks it's been written to
> venti, then fossil would have no access without a venti server.
>
> so it would seem to me that the simplist explination for this is
> that although you think you haven't configured venti, somehow
> it was.
>
> do you have a venti paritition?  also the output of venti/conf
> would be helpful.
>
> - erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[john]

On Sun Dec 16 20:58:30 EST pietro10@mac.com wrote:
> I have no venti partition, and venti/conf asks for one called  
> v.arenas. The reason was that the default configuration of the  
> installer only gave fossil 700MB and venti/arenas 3GB. I wasn't sure  
> that would be sufficient for my files :-)
> 

Unless I completely misunderstand the way venti and fossil work, that
would be quite sufficient unless you plan on creating more than
about 700 MB of files in a day. After you install, fossil will flush everything
to venti, which *should* get you a pretty much empty fossil partition--
it's just a cache, remember. Your files will then be moved from fossil
to venti every night thereafter. Since venti practices block compression,
it shouldn't be much of a problem to have venti... unless you copy in
a whole bunch of mp3 files and then decide you don't want them *after*
venti has written them.
I used a fossil+venti system to store a bunch of music at one point. I would
fill up the fossil buffer with mp3s, then force a sync to venti, then re-fill
the fossil buffer, repeating until I had transferred everything and written
it to venti.
If the things I've said here are wrong or don't make sense, let me know.

John

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

Then where are the nodes stored?

On Dec 17, 2007, at 2:24 PM, john@csplan9.rit.edu wrote:

> On Sun Dec 16 20:58:30 EST pietro10@mac.com wrote:
>> I have no venti partition, and venti/conf asks for one called
>> v.arenas. The reason was that the default configuration of the
>> installer only gave fossil 700MB and venti/arenas 3GB. I wasn't sure
>> that would be sufficient for my files :-)
>>
>
> Unless I completely misunderstand the way venti and fossil work, that
> would be quite sufficient unless you plan on creating more than
> about 700 MB of files in a day. After you install, fossil will  
> flush everything
> to venti, which *should* get you a pretty much empty fossil  
> partition--
> it's just a cache, remember. Your files will then be moved from fossil
> to venti every night thereafter. Since venti practices block  
> compression,
> it shouldn't be much of a problem to have venti... unless you copy in
> a whole bunch of mp3 files and then decide you don't want them *after*
> venti has written them.
> I used a fossil+venti system to store a bunch of music at one  
> point. I would
> fill up the fossil buffer with mp3s, then force a sync to venti,  
> then re-fill
> the fossil buffer, repeating until I had transferred everything and  
> written
> it to venti.
> If the things I've said here are wrong or don't make sense, let me  
> know.
>
> John
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[erik quanstrom]

> Then where are the nodes stored?
> 
> On Dec 17, 2007, at 2:24 PM, john@csplan9.rit.edu wrote:
>
[...]
> >
> > Unless I completely misunderstand the way venti and fossil work, that
> > would be quite sufficient unless you plan on creating more than
> > about 700 MB of files in a day. After you install, fossil will  
> > flush everything
> > to venti, which *should* get you a pretty much empty fossil  
> > partition--
> > it's just a cache, remember. Your files will then be moved from fossil
> > to venti every night thereafter. Since venti practices block  
> > compression,
> > it shouldn't be much of a problem to have venti... unless you copy in
> > a whole bunch of mp3 files and then decide you don't want them *after*
> > venti has written them.
> > I used a fossil+venti system to store a bunch of music at one  
> > point. I would
> > fill up the fossil buffer with mp3s, then force a sync to venti,  
> > then re-fill
> > the fossil buffer, repeating until I had transferred everything and  
> > written
> > it to venti.
> > If the things I've said here are wrong or don't make sense, let me  
> > know.
> >
> > John

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[erik quanstrom]

wierd.  that's not what i typed.

anyway, if by nodes you mean directory entries, these are
stored in the same manner as the data.  /sys/doc/fossil.ps
has the details.

- erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I'd still like to know how to check my file system without dealing  
with fsOpen, if that is possible. Does anyone know how?

On Dec 16, 2007, at 8:42 PM, erik quanstrom wrote:

> On Sun Dec 16 20:37:03 EST 2007, pietro10@mac.com wrote:
>> And now it still says it can't open the file system. What's wrong
>> with checking the raw data? I don't want to lose my files, and
>> nothing seems to work without calling fsOpen.
>
> fossil is just a write buffer.  if fossil thinks it's been written to
> venti, then fossil would have no access without a venti server.
>
> so it would seem to me that the simplist explination for this is
> that although you think you haven't configured venti, somehow
> it was.
>
> do you have a venti paritition?  also the output of venti/conf
> would be helpful.
>
> - erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Juan M. Mendez]

On 19/12/2007, Pietro Gagliardi <pietro10@mac.com> wrote:
> I'd still like to know how to check my file system without dealing
> with fsOpen, if that is possible. Does anyone know how?


I posted this in another topic, maybe it gives you ideas of how to
deal with your problem.
Whenever I have problems with my disk (a real old bad disk) I do this to get my
fossil only file system fixed:

<<
I booted with the plan9 livecd and did a check to the disk with this.

fossil/fossil -c 'srv -p fscons'
con /srv/fscons
prompt: srv -AWP replica
prompt: fsys main config /dev/sdC0/fossil
prompt: fsys main open -AWP
warning: connecting to venti: cs: can't translate address: '/srv/dns' file
does not exist
prompt: fsys main
main: check fix

>>
-- 
Fidonet: 2:345/432.2

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I'm sorry, I still get the errors. I'll print them verbatim:

	fsOpen error
	fsOpen: fileRoot: corrupted file entry

No matter how I word open, this error comes up. My question: why not  
just check fix the raw data?

PS - The -V option to open can be used to avoid venti checking.

On Dec 19, 2007, at 8:15 PM, Juan M. Mendez wrote:

> On 19/12/2007, Pietro Gagliardi <pietro10@mac.com> wrote:
>> I'd still like to know how to check my file system without dealing
>> with fsOpen, if that is possible. Does anyone know how?
>
>
> I posted this in another topic, maybe it gives you ideas of how to
> deal with your problem.
> Whenever I have problems with my disk (a real old bad disk) I do  
> this to get my
> fossil only file system fixed:
>
> <<
> I booted with the plan9 livecd and did a check to the disk with this.
>
> fossil/fossil -c 'srv -p fscons'
> con /srv/fscons
> prompt: srv -AWP replica
> prompt: fsys main config /dev/sdC0/fossil
> prompt: fsys main open -AWP
> warning: connecting to venti: cs: can't translate address: '/srv/ 
> dns' file
> does not exist
> prompt: fsys main
> main: check fix
>
>>>
> -- 
> Fidonet: 2:345/432.2

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[erik quanstrom]

> 	fsOpen error
> 	fsOpen: fileRoot: corrupted file entry
> 
> No matter how I word open, this error comes up. My question: why not  
> just check fix the raw data?

fossil can't open the root and can't help you.  i've suggested using
fossil/flchk but you don't want to do that.  i think that's as close
to "checking the raw data" as you can get.  why don't you want
to use the tool available?

- erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I did do flchk but it gives me the same error >:-) Everything is  
failing!

On Dec 19, 2007, at 9:02 PM, erik quanstrom wrote:

>> 	fsOpen error
>> 	fsOpen: fileRoot: corrupted file entry
>>
>> No matter how I word open, this error comes up. My question: why not
>> just check fix the raw data?
>
> fossil can't open the root and can't help you.  i've suggested using
> fossil/flchk but you don't want to do that.  i think that's as close
> to "checking the raw data" as you can get.  why don't you want
> to use the tool available?
>
> - erik

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

Okay, so is there any tool available that can try to extract my files  
without going through fsOpen? flchk refuses to fix it, so I guess I'm  
screwed. :-( And I don't know what caused this!

On Dec 19, 2007, at 9:10 PM, Pietro Gagliardi wrote:

> I did do flchk but it gives me the same error >:-) Everything is  
> failing!
>
> On Dec 19, 2007, at 9:02 PM, erik quanstrom wrote:
>
>>> 	fsOpen error
>>> 	fsOpen: fileRoot: corrupted file entry
>>>
>>> No matter how I word open, this error comes up. My question: why not
>>> just check fix the raw data?
>>
>> fossil can't open the root and can't help you.  i've suggested using
>> fossil/flchk but you don't want to do that.  i think that's as close
>> to "checking the raw data" as you can get.  why don't you want
>> to use the tool available?
>>
>> - erik
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

Can someone PLEASE give me some information so I can get my files  
back? Otherwise, there goes my 3D library, my hoc implementation, etc.

On Dec 20, 2007, at 2:58 PM, Pietro Gagliardi wrote:

> Okay, so is there any tool available that can try to extract my  
> files without going through fsOpen? flchk refuses to fix it, so I  
> guess I'm screwed. :-( And I don't know what caused this!
>
> On Dec 19, 2007, at 9:10 PM, Pietro Gagliardi wrote:
>
>> I did do flchk but it gives me the same error >:-) Everything is  
>> failing!
>>
>> On Dec 19, 2007, at 9:02 PM, erik quanstrom wrote:
>>
>>>> 	fsOpen error
>>>> 	fsOpen: fileRoot: corrupted file entry
>>>>
>>>> No matter how I word open, this error comes up. My question: why  
>>>> not
>>>> just check fix the raw data?
>>>
>>> fossil can't open the root and can't help you.  i've suggested using
>>> fossil/flchk but you don't want to do that.  i think that's as close
>>> to "checking the raw data" as you can get.  why don't you want
>>> to use the tool available?
>>>
>>> - erik
>>
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Bakul Shah]

> Can someone PLEASE give me some information so I can get my files  
> back? Otherwise, there goes my 3D library, my hoc implementation, etc.
> 
> On Dec 20, 2007, at 2:58 PM, Pietro Gagliardi wrote:
> 
> > Okay, so is there any tool available that can try to extract my  
> > files without going through fsOpen? flchk refuses to fix it, so I  
> > guess I'm screwed. :-( And I don't know what caused this!

One idea is to install plan9 from scratch on a different qemu
image and give this image to qemu as the second disk (option
-hdb).  If the venti arenas are in mostly good shape you can
try playing with venti/wrarena.  You can probably try this
from your host OS using p9p venti.  Or you can write ventidb
(ala fsdb for debugging FFS on *BSD) that will allow you to
interactively navigate venti.  And may be something similar
for fossil?  No idea if any of this would work but this will
keep you busy for a while.

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Russ Cox]

> Can someone PLEASE give me some information so I can get my files  
> back? Otherwise, there goes my 3D library, my hoc implementation, etc.

I'm afraid you're probably out of luck, without a lot of work.

The disk data structures are all described in /sys/doc/fossil.ps.

If you were using snapshots, then you might be able to 
use the information in /sys/doc/fossil.ps to write a program
to find all the fossil snapshot roots, and then you 
could rewrite the super block to point at an older, 
hopefully not corrupt, root.

However, it looks like in plain fossil mode (not fossil+venti),
the installer does not configure any snapshots, so you'd
have had to turn them on yourself.

Without snapshots, you'd have to poke around on the
disk looking for a VtSource that looked like the root
directory and then fix up the root block to be valid and
point at it.

If you're not used to poking at file system data structures,
this might well be more work than just recreating your
hoc and 3D code.

Russ

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

OK, I'm going to hold this off for the new year. Until then, I  
managed to free up a lot of space (and am continuing to do so), so  
this time I'll put up a new 4GB QEMU fossil+venti Plan 9 and save the  
old disk until I can fix it.

On Dec 27, 2007, at 11:23 AM, Russ Cox wrote:

>> Can someone PLEASE give me some information so I can get my files
>> back? Otherwise, there goes my 3D library, my hoc implementation,  
>> etc.
>
> I'm afraid you're probably out of luck, without a lot of work.
>
> The disk data structures are all described in /sys/doc/fossil.ps.
>
> If you were using snapshots, then you might be able to
> use the information in /sys/doc/fossil.ps to write a program
> to find all the fossil snapshot roots, and then you
> could rewrite the super block to point at an older,
> hopefully not corrupt, root.
>
> However, it looks like in plain fossil mode (not fossil+venti),
> the installer does not configure any snapshots, so you'd
> have had to turn them on yourself.
>
> Without snapshots, you'd have to poke around on the
> disk looking for a VtSource that looked like the root
> directory and then fix up the root block to be valid and
> point at it.
>
> If you're not used to poking at file system data structures,
> this might well be more work than just recreating your
> hoc and 3D code.
>
> Russ
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

Right now, I'm trying to get function definitions off of the plan9  
file so I can rebuild my profile. This should be as easy as grep.

On Dec 29, 2007, at 11:33 AM, Pietro Gagliardi wrote:

> OK, I'm going to hold this off for the new year. Until then, I  
> managed to free up a lot of space (and am continuing to do so), so  
> this time I'll put up a new 4GB QEMU fossil+venti Plan 9 and save  
> the old disk until I can fix it.
>
> On Dec 27, 2007, at 11:23 AM, Russ Cox wrote:
>
>>> Can someone PLEASE give me some information so I can get my files
>>> back? Otherwise, there goes my 3D library, my hoc implementation,  
>>> etc.
>>
>> I'm afraid you're probably out of luck, without a lot of work.
>>
>> The disk data structures are all described in /sys/doc/fossil.ps.
>>
>> If you were using snapshots, then you might be able to
>> use the information in /sys/doc/fossil.ps to write a program
>> to find all the fossil snapshot roots, and then you
>> could rewrite the super block to point at an older,
>> hopefully not corrupt, root.
>>
>> However, it looks like in plain fossil mode (not fossil+venti),
>> the installer does not configure any snapshots, so you'd
>> have had to turn them on yourself.
>>
>> Without snapshots, you'd have to poke around on the
>> disk looking for a VtSource that looked like the root
>> directory and then fix up the root block to be valid and
>> point at it.
>>
>> If you're not used to poking at file system data structures,
>> this might well be more work than just recreating your
>> hoc and 3D code.
>>
>> Russ
>>
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I decided to take a peek at the corrupted entry. All of the fields of  
the Entry are zeros! This causes the sourceAlloc() function to fail,  
and thus I can't get a source without the program crashing (I tried  
once). Is there a way to fix this?

On Dec 27, 2007, at 11:23 AM, Russ Cox wrote:

>> Can someone PLEASE give me some information so I can get my files
>> back? Otherwise, there goes my 3D library, my hoc implementation,  
>> etc.
>
> I'm afraid you're probably out of luck, without a lot of work.
>
> The disk data structures are all described in /sys/doc/fossil.ps.
>
> If you were using snapshots, then you might be able to
> use the information in /sys/doc/fossil.ps to write a program
> to find all the fossil snapshot roots, and then you
> could rewrite the super block to point at an older,
> hopefully not corrupt, root.
>
> However, it looks like in plain fossil mode (not fossil+venti),
> the installer does not configure any snapshots, so you'd
> have had to turn them on yourself.
>
> Without snapshots, you'd have to poke around on the
> disk looking for a VtSource that looked like the root
> directory and then fix up the root block to be valid and
> point at it.
>
> If you're not used to poking at file system data structures,
> this might well be more work than just recreating your
> hoc and 3D code.
>
> Russ
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I decided to stop doing this with Plan 9 and decided to do it from  
within Mac OS X. I'm running Plan 9 atop QEMU, which is implemented  
as a program called Q (http://www.kju-app.org/). The first step was  
to convert the compressed QCOW hard disk image to a raw one for  
analysis:

	$ /Applications/Q.app/Contents/MacOS/qemu-img convert ~/Documents/ 
QEMU/Plan\ 9\ from\ Bell\ Labs.qvm/Harddisk_1.qcow2 -O raw ~/Desktop/ 
hd.raw

When that is done, I'll examine hd.raw in a text editor and get my  
files back.

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

I got that done and did a test run. I found a book I was working on! :-)

On Jan 11, 2008, at 3:01 PM, Pietro Gagliardi wrote:

> I decided to stop doing this with Plan 9 and decided to do it from  
> within Mac OS X. I'm running Plan 9 atop QEMU, which is implemented  
> as a program called Q (http://www.kju-app.org/). The first step was  
> to convert the compressed QCOW hard disk image to a raw one for  
> analysis:
>
> 	$ /Applications/Q.app/Contents/MacOS/qemu-img convert ~/Documents/ 
> QEMU/Plan\ 9\ from\ Bell\ Labs.qvm/Harddisk_1.qcow2 -O raw ~/ 
> Desktop/hd.raw
>
> When that is done, I'll examine hd.raw in a text editor and get my  
> files back.
>

Re: [9fans] Corrupted file entry on QEMU - how to recover?

[Pietro Gagliardi]

That part was missing some lines, so I decided to start over. This  
time, an AWK script is used:

	awk '
	/^fn unt[gb]z/ { p(); next; pr(); }
	/^fn df/ { pr(); next; pr(); }
	/^\.C$/ { pr() }
	function pr() { print NR, $0 }' hd.raw

untgz and untbz are functions that extract .tar.gz/.tgz  
and .tar.bz2/.tbz files automatically. df shows the disk usage. .C is  
the macro in my book for a chapter (one of the lines that was  
missing). I tell AWK to print the multi-line functions (with line  
numbers).

On Jan 11, 2008, at 3:47 PM, Pietro Gagliardi wrote:

> I got that done and did a test run. I found a book I was working  
> on! :-)
>
> On Jan 11, 2008, at 3:01 PM, Pietro Gagliardi wrote:
>
>> I decided to stop doing this with Plan 9 and decided to do it from  
>> within Mac OS X. I'm running Plan 9 atop QEMU, which is  
>> implemented as a program called Q (http://www.kju-app.org/). The  
>> first step was to convert the compressed QCOW hard disk image to a  
>> raw one for analysis:
>>
>> 	$ /Applications/Q.app/Contents/MacOS/qemu-img convert ~/Documents/ 
>> QEMU/Plan\ 9\ from\ Bell\ Labs.qvm/Harddisk_1.qcow2 -O raw ~/ 
>> Desktop/hd.raw
>>
>> When that is done, I'll examine hd.raw in a text editor and get my  
>> files back.
>>
>