VirtualHwStorFindAdapter after HwStorFreeAdapterResources

cristalink

unread,

Jan 2, 2010, 4:00:04 AM1/2/10

to

Hello,

I have a memory dump from Microsoft OCA that indicates that the system
called MyDriver!VirtualHwStorFindAdapter() after
MyDriver!HwStorFreeAdapterResources() in my virtual StorPort miniport. My
driver passed Winqual tests before. I tried sleep, hibernation, and "devcon
restart mydriver", but I couldn't reproduce the call sequence. How can I
reproduce the issue? Under which circumstances StorPort "restarts" my device
and calls VirtualHwStorFindAdapter() without unloading my driver first after
HwStorFreeAdapterResources()?

Thanks,

Nick

STACK_TEXT:
f78ea97c b865dfee 89e5b3dc 8948b030 89e6c730
MyDriver!VirtualHwStorFindAdapter+0x538
f78ea9b0 b865a0b6 00e6c8a0 b865df6c 00000000
storport!RaCallMiniportFindAdapter+0x7c
f78ea9d4 b868abce 89e6c7e8 89b66700 00000000
storport!RaidAdapterStartMiniport+0x5a
f78ea9f4 b868ad64 e138cec8 89b666a8 b8668510
storport!RaidAdapterStartDeviceIrp+0xb0
f78eaa14 b868b24f 01e6c7e8 00000000 89b66760 storport!RaidAdapterPnpIrp+0xee
f78eaa30 80840153 89e6c730 89b666a8 f78eaab0 storport!RaDriverPnpIrp+0x6d
f78eaa44 80901c8f f78eaab0 8948b030 00000000 nt!IofCallDriver+0x45
f78eaa70 808de957 89e6c730 f78eaa8c 00000000 nt!IopSynchronousCall+0xb8
f78eaab4 808deac4 8948b030 89cdad58 00000001 nt!IopStartDevice+0x4d
f78eaad0 808dea89 8948b030 8948b001 89cdad58 nt!PipProcessStartPhase1+0x4e
f78ead28 808bc4de 89444008 00000001 00000000 nt!PipProcessDevNodeTree+0x1db
f78ead58 80801be5 00000003 89f97660 808b711c nt!PiRestartDevice+0x80
f78ead80 8082db10 00000000 00000000 89f97660 nt!PipDeviceActionWorker+0x188
f78eadac 8092083b 00000000 00000000 00000000 nt!ExpWorkerThread+0xeb
f78eaddc 8083fe9f 8082da53 00000001 00000000 nt!PspSystemThreadStartup+0x2e
00000000 00000000 00000000 00000000 00000000 nt!KiThreadStartup+0x16

Alexander Grigoriev

unread,

Jan 2, 2010, 3:35:45 PM1/2/10

to

Are you saying that you haven't received SRB_FUNCTION_PNP/StorRemoveDevice
for it?

I saw such restart if DO_POWER_PAGABLE flag was set in the PDO. If I
remember correctly.

"cristalink" <crist...@noemail.noemail> wrote in message
news:exJb8n4i...@TK2MSFTNGP04.phx.gbl...

cristalink

unread,

Jan 2, 2010, 4:31:29 PM1/2/10

to

In my current design, DriverEntry allocates memory and stores the pointer in
a global variable. VirtualHwStorFindAdapter then uses the global pointer.
HwStorFreeAdapterResources frees the memory and zeroes the pointer. If
VirtualHwStorFindAdapter is called again, it uses the null pointer, hence
the crash.

I cannot repro it. All I have is a small minidump with the call stack and
registers. StorPort is almost undocumented, so I can't fix the issue without
being able to repro it.

> Are you saying that you haven't received SRB_FUNCTION_PNP/StorRemoveDevice
> for it?

I have the call stack only, so I cannot say whether StorRemoveDevice was
called or not. But it shouldn't matter anyway. All I need is to repro the
case.

Nick

"Alexander Grigoriev" <al...@earthlink.net> wrote in message
news:%23SqVqs%23iKH...@TK2MSFTNGP04.phx.gbl...

Alexander Grigoriev

unread,

Jan 2, 2010, 8:35:23 PM1/2/10

to

There is no guarantee whatsoever the driver binary will be reloaded every
time before FindAdapter is called. Even for a root-enumerated device, which
you probably have. For example, there is storport.sys bug, which leaves a
hanging device object every time a SCSI target disappears from under it. If
that happens, your binary will bever be unloaded until a reboot.

"cristalink" <crist...@noemail.noemail> wrote in message

news:O7IE2L$iKHA...@TK2MSFTNGP06.phx.gbl...