restoring pci config space for a device

Hi.

a device (a DSP) that my driver handles has somekind of a reset process.
in this reset process the pci configuration space on the device, which was
written to the device by the OS on startup, is deleted and it is my
responsibility to restore it since the device should continue functioning
without restarting the machine (which is the only point of time where the OS
writes the pci config values to my understanding).

my question is if there's a common/acceptable way of doing so.

my cuurent thought will be -
keeping an instance of PCI_COMMON_CONFIG which was read by
irp_mn_read_config and use the irp_mn_write_config for restoring.

thanks a lot,
Kobi.

You can implement reset as the driver's start/stop cycle.

And, if the device resets its PCI interface due to some device-side events,
then it is _just plain buggy_.

--
Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation

http://www.storagecraft.com

"kobi n" <> wrote in message
news:8D27D94F-DDFA-4ED2-8532-...
> Hi.
>
> a device (a DSP) that my driver handles has somekind of a reset process.
> in this reset process the pci configuration space on the device, which was
> written to the device by the OS on startup, is deleted and it is my
> responsibility to restore it since the device should continue functioning
> without restarting the machine (which is the only point of time where the OS
> writes the pci config values to my understanding).
>
> my question is if there's a common/acceptable way of doing so.
>
> my cuurent thought will be -
> keeping an instance of PCI_COMMON_CONFIG which was read by
> irp_mn_read_config and use the irp_mn_write_config for restoring.
>
> thanks a lot,
> Kobi.
>
>
>

kobi n <> wrote:
>
>a device (a DSP) that my driver handles has somekind of a reset process.
>in this reset process the pci configuration space on the device, which was
>written to the device by the OS on startup, is deleted and it is my
>responsibility to restore it since the device should continue functioning
>without restarting the machine (which is the only point of time where the OS
>writes the pci config values to my understanding).
>
>my question is if there's a common/acceptable way of doing so.

No, because your hardware is broken. This behavior violates the PCI
specification.
--
Tim Roberts,
Providenza & Boekelheide, Inc.

yes. when your device resets, try to disable / enable it
using devcon.exe.

/* Maxim: you know, any hardware bug fixable in the sofware,
is a software bug ... */

--PA

"Maxim S. Shatskih" <> wrote in message news:...
> You can implement reset as the driver's start/stop cycle.
>
> And, if the device resets its PCI interface due to some device-side events,
> then it is _just plain buggy_.
>
> --
> Maxim Shatskih, Windows DDK MVP
> StorageCraft Corporation
>
> http://www.storagecraft.com
>
> "kobi n" <> wrote in message
> news:8D27D94F-DDFA-4ED2-8532-...
>> Hi.
>>
>> a device (a DSP) that my driver handles has somekind of a reset process.
>> in this reset process the pci configuration space on the device, which was
>> written to the device by the OS on startup, is deleted and it is my
>> responsibility to restore it since the device should continue functioning
>> without restarting the machine (which is the only point of time where the OS
>> writes the pci config values to my understanding).
>>
>> my question is if there's a common/acceptable way of doing so.
>>
>> my cuurent thought will be -
>> keeping an instance of PCI_COMMON_CONFIG which was read by
>> irp_mn_read_config and use the irp_mn_write_config for restoring.
>>
>> thanks a lot,
>> Kobi.
>>
>>
>>
>

sorry, i did not understand.
what do i earn by disable/enable using the devcon ? does the OS rewrites
pci config info ?

"Pavel A." wrote:

> yes. when your device resets, try to disable / enable it
> using devcon.exe.
>
> /* Maxim: you know, any hardware bug fixable in the sofware,
> is a software bug ... */
>
> --PA
>
> "Maxim S. Shatskih" <> wrote in message news:...
> > You can implement reset as the driver's start/stop cycle.
> >
> > And, if the device resets its PCI interface due to some device-side events,
> > then it is _just plain buggy_.
> >
> > --
> > Maxim Shatskih, Windows DDK MVP
> > StorageCraft Corporation
> >
> > http://www.storagecraft.com
> >
> > "kobi n" <> wrote in message
> > news:8D27D94F-DDFA-4ED2-8532-...
> >> Hi.
> >>
> >> a device (a DSP) that my driver handles has somekind of a reset process.
> >> in this reset process the pci configuration space on the device, which was
> >> written to the device by the OS on startup, is deleted and it is my
> >> responsibility to restore it since the device should continue functioning
> >> without restarting the machine (which is the only point of time where the OS
> >> writes the pci config values to my understanding).
> >>
> >> my question is if there's a common/acceptable way of doing so.
> >>
> >> my cuurent thought will be -
> >> keeping an instance of PCI_COMMON_CONFIG which was read by
> >> irp_mn_read_config and use the irp_mn_write_config for restoring.
> >>
> >> thanks a lot,
> >> Kobi.
> >>
> >>
> >>
> >
>
>
>

ok. i understand the problem.
but i have a very difficult time understanding the "IRP_MN_WRITE_CONFIG".

i can't seem to perform any writes with that. i can't even find a single
example.

assuming i used IRP_MN_WRITE_CONFIG to obtain a device's configuratino space
and hold it in a PCI_COMMON_CONFIG struct. i then change , for example,
the value of VendorId in it. then i try to write it again using the
"IRP_MN_WRITE_CONFIG" but nothing seems to happen. on the next READ i get
the same old value.

any idea why can't i perform the write ??



"Maxim S. Shatskih" wrote:

> You can implement reset as the driver's start/stop cycle.
>
> And, if the device resets its PCI interface due to some device-side events,
> then it is _just plain buggy_.
>
> --
> Maxim Shatskih, Windows DDK MVP
> StorageCraft Corporation
>
> http://www.storagecraft.com
>
> "kobi n" <> wrote in message
> news:8D27D94F-DDFA-4ED2-8532-...
> > Hi.
> >
> > a device (a DSP) that my driver handles has somekind of a reset process.
> > in this reset process the pci configuration space on the device, which was
> > written to the device by the OS on startup, is deleted and it is my
> > responsibility to restore it since the device should continue functioning
> > without restarting the machine (which is the only point of time where the OS
> > writes the pci config values to my understanding).
> >
> > my question is if there's a common/acceptable way of doing so.
> >
> > my cuurent thought will be -
> > keeping an instance of PCI_COMMON_CONFIG which was read by
> > irp_mn_read_config and use the irp_mn_write_config for restoring.
> >
> > thanks a lot,
> > Kobi.
> >
> >
> >
>
>

isn't using IRP_MN_WRITE_CONFIG and IRP_MN_READ_CONFIG is fully equivalent
to the obsolete HalSetBusData & HalGetBusData ?

it is described so in the documentation.
BUT. using the Hal routines in an old driver i was able to perform both
read and write, now using the IRP_MN_WRITE_CONFIG doesnt seem to work at
all


"Maxim S. Shatskih" wrote:

> You can implement reset as the driver's start/stop cycle.
>
> And, if the device resets its PCI interface due to some device-side events,
> then it is _just plain buggy_.
>
> --
> Maxim Shatskih, Windows DDK MVP
> StorageCraft Corporation
>
> http://www.storagecraft.com
>
> "kobi n" <> wrote in message
> news:8D27D94F-DDFA-4ED2-8532-...
> > Hi.
> >
> > a device (a DSP) that my driver handles has somekind of a reset process.
> > in this reset process the pci configuration space on the device, which was
> > written to the device by the OS on startup, is deleted and it is my
> > responsibility to restore it since the device should continue functioning
> > without restarting the machine (which is the only point of time where the OS
> > writes the pci config values to my understanding).
> >
> > my question is if there's a common/acceptable way of doing so.
> >
> > my cuurent thought will be -
> > keeping an instance of PCI_COMMON_CONFIG which was read by
> > irp_mn_read_config and use the irp_mn_write_config for restoring.
> >
> > thanks a lot,
> > Kobi.
> >
> >
> >
>
>

> and hold it in a PCI_COMMON_CONFIG struct. i then change , for example,
> the value of VendorId in it. then i try to write it again using the

If you change the vendor ID, then you change the PnP ID of the device, which
means that your driver is gone and another driver will be loaded, so, you will
not be able to restore the config space.

--
Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation

http://www.storagecraft.com

"kobi n" <> wrote in message news:5BEAFD38-210C-4DFB-A47C-...
>
> sorry, i did not understand.
> what do i earn by disable/enable using the devcon ? does the OS rewrites
> pci config info ?

There is a chance that after re-enabling, the OS ( = the PCI bus driver) would
re-initialize the config info of your card, so your driver will restart and go on.
Of course this chance depends on how bad is the h/w bug
(are the vendor/device ID regs etc. erased?) but this shoud be easy to test.

Regards,
--PA

>
> "Pavel A." wrote:
>
>> yes. when your device resets, try to disable / enable it
>> using devcon.exe.
>>
>> /* Maxim: you know, any hardware bug fixable in the sofware,
>> is a software bug ... */
>>
>> --PA
>>
>> "Maxim S. Shatskih" <> wrote in message news:...
>> > You can implement reset as the driver's start/stop cycle.
>> >
>> > And, if the device resets its PCI interface due to some device-side events,
>> > then it is _just plain buggy_.
>> >
>> > --
>> > Maxim Shatskih, Windows DDK MVP
>> > StorageCraft Corporation
>> >
>> > http://www.storagecraft.com
>> >
>> > "kobi n" <> wrote in message
>> > news:8D27D94F-DDFA-4ED2-8532-...
>> >> Hi.
>> >>
>> >> a device (a DSP) that my driver handles has somekind of a reset process.
>> >> in this reset process the pci configuration space on the device, which was
>> >> written to the device by the OS on startup, is deleted and it is my
>> >> responsibility to restore it since the device should continue functioning
>> >> without restarting the machine (which is the only point of time where the OS
>> >> writes the pci config values to my understanding).
>> >>
>> >> my question is if there's a common/acceptable way of doing so.
>> >>
>> >> my cuurent thought will be -
>> >> keeping an instance of PCI_COMMON_CONFIG which was read by
>> >> irp_mn_read_config and use the irp_mn_write_config for restoring.
>> >>
>> >> thanks a lot,
>> >> Kobi.
>> >>
>> >>
>> >>
>> >
>>
>>
>>

kobi n <> wrote:
>
>isn't using IRP_MN_WRITE_CONFIG and IRP_MN_READ_CONFIG is fully equivalent
>to the obsolete HalSetBusData & HalGetBusData ?

Not fully, no. HalSetBusData can write to arbitrary PCI devices, by
specifying bus/slot/function numbers. IRP_MN_WRITE_CONFIG can only affect
the device that has been assigned to your driver.
--
Tim Roberts,
Providenza & Boekelheide, Inc.