!vm output

WINDBG
1

We have a situation where we are apparently running a machine out of
memory with our filesystem code. While trying to depug this, we see is
ResAvail pages (displayed from !vm) has gone to 317 (on a 2GB machine).

The question here is “What is RedAvail pages”? Resident, Resource, ??
And what is this very low number telling us?

(I’ve looked at the debugger docs and Windows Internals books and don’t
find a definitive answer here…)

Thanks…dave

2

Here’s the response from the MM expert:

This is the resident available pages - the number of free pages there
would be if everything that needs to be resident in ram actually was
present at the same time (ie, all kernel thread stacks have been swapped
in, all processes are at their working set minimums, etc).

The 317 value is very low.

It likely indicates that they have some process who calls
SetProcessWorkingSetSize{Ex} to raise its minimum working set (!process
0 1 will tell you if this is the case) or lots of threads (each has a
kernel stack cost) or they are neglecting to unlock MDLs, etc.

If you give me the exact OS release and a dump of the first 0x80
ULONGS_PTRs in nt!mmrestrack I might be able to narrow down a bit for
them.

-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Dave Beaver
Sent: Tuesday, December 06, 2005 10:23 AM
To: Kernel Debugging Interest List
Subject: [windbg] !vm output

We have a situation where we are apparently running a machine out of
memory with our filesystem code. While trying to depug this, we see is
ResAvail pages (displayed from !vm) has gone to 317 (on a 2GB machine).

The question here is “What is RedAvail pages”? Resident, Resource, ??
And what is this very low number telling us?

(I’ve looked at the debugger docs and Windows Internals books and don’t
find a definitive answer here…)

Thanks…dave

You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com

3

Thanks for the quick answer. Turns out what was running at the time we
experienced this was the WHQL HCT iostress test - doing exactly what
it’s supposed to do, comsuming memory and seeing how the FS deals with
memory allocation failures.

Thanks!..dave

-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Drew Bliss
Sent: Tuesday, December 06, 2005 4:05 PM
To: Kernel Debugging Interest List
Subject: RE: [windbg] !vm output

Here’s the response from the MM expert:

This is the resident available pages - the number of free pages there
would be if everything that needs to be resident in ram actually was
present at the same time (ie, all kernel thread stacks have been swapped
in, all processes are at their working set minimums, etc).

The 317 value is very low.

It likely indicates that they have some process who calls
SetProcessWorkingSetSize{Ex} to raise its minimum working set (!process
0 1 will tell you if this is the case) or lots of threads (each has a
kernel stack cost) or they are neglecting to unlock MDLs, etc.

If you give me the exact OS release and a dump of the first 0x80
ULONGS_PTRs in nt!mmrestrack I might be able to narrow down a bit for
them.

-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Dave Beaver
Sent: Tuesday, December 06, 2005 10:23 AM
To: Kernel Debugging Interest List
Subject: [windbg] !vm output

We have a situation where we are apparently running a machine out of
memory with our filesystem code. While trying to depug this, we see is
ResAvail pages (displayed from !vm) has gone to 317 (on a 2GB machine).

The question here is “What is RedAvail pages”? Resident, Resource, ??
And what is this very low number telling us?

(I’ve looked at the debugger docs and Windows Internals books and don’t
find a definitive answer here…)

Thanks…dave

You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com

You are currently subscribed to windbg as: unknown lmsubst tag argument:
‘’
To unsubscribe send a blank email to xxxxx@lists.osr.com