You have to return a value that is a warning, not a failure, to copy back data into the output buffer
C:\Program Files (x86)\Windows Kits\8.0\Include\shared>findstr STATUS_BUFFER_TOO_SMALL ntstatus.h
// MessageId: STATUS_BUFFER_TOO_SMALL
#define STATUS_BUFFER_TOO_SMALL ((NTSTATUS)0xC0000023L) <– error
C:\Program Files (x86)\Windows Kits\8.0\Include\shared>findstr STATUS_BUFFER_OVERFLOW ntstatus.h
// MessageId: STATUS_BUFFER_OVERFLOW
#define STATUS_BUFFER_OVERFLOW ((NTSTATUS)0x80000005L) <– warning
From ntstatus.h, explaining the top two bits
// Values are 32 bit values laid out as follows:
//
// 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
// 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
// ±–±±±----------------------±------------------------------+
// |Sev|C|R| Facility | Code |
// ±–±±±----------------------±------------------------------+
//
// where
//
// Sev - is the severity code
//
// 00 - Success
// 01 - Informational
// 10 - Warning
// 11 - Error
-----Original Message-----
From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@flounder.com
Sent: Monday, September 17, 2012 7:56 PM
To: Windows System Software Devs Interest List
Subject: RE: [ntdev] Passing variable-sized array from Kernel to User
There are several possible approaches, but there’s a bit of a glitch from drivers. The problem I found was that if I returned an error code, like buffer-too-small, but had filled in as much of the buffer as I had, in buffered mode the buffer was not copied back, and in direct mode it returned 0 for bytes-transferred
So the one way I know that will work is the following:
BOOL ok;
SIZE_T size;
size = some_initial_value_of_your_choice;
WHATEVER * buffer = new WHATEVER[size];
while(TRUE)
{ /* look */
ok = DeviceIoControl( …size… &bytesTrannsferred …);
if(!ok)
{ /* failed */
DWORD err = ::GetLastError();
if(err == ERROR_WHATEVER) // buffer too small
{ /* buffer too small */
size * = 2;
delete buffer;
buffer = new WHATEVER[size];
continue;
} /* buffer too small */
break; // exit loop
} /* failed */
else
{ /* no error */
if(bytesTransferred == size)
{ /* exact match */
size *= 2;
delete buffer;
buffer = new WHATEVER[size];
continue;
} /* exact match */
} /* no error */
break;
} /* loop */
if(!ok)
{ /* solid error */
…deal with it
} /* solid error */
That is, if you fill the buffer exactly, you might have lost data, so the right thing to do is to try again with a larger buffer, and continue until you get an only-partially-filled buffer. There are several user-level APIs for which this is the only possible solution.
Note the code above can be refactored to avoid duplication, but I’m typing this in without a decent program editor, so I didn’t bother to reduce it, as I would have done had I been coding it in an actual deliverable product. Also, in writing C++, I would more likely use a collection like CArray or std::vector, which is why I’m a bit sketchy about details like how to delete and reallocate the buffer I would not use delete and new as shown, but something more complex but more robust.
joe
> The problem with the two ioctl approach (get size, then get buffer) is
> the window between the two. The size can change in which case you have
> to indicate “there is more” somehow. With the get buffer alone, you
> have the same issue of how to indicate there is more. So, since they
> both devolve to the same problem, just have a single get buffer ioctl.
> The driver can choose to fill in as many slots in the array as it has
> data
>
> d
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of JonathonS
> Sent: Monday, September 17, 2012 5:30 PM
> To: Windows System Software Devs Interest List
> Subject: [ntdev] Passing variable-sized array from Kernel to User
>
> Hi all,
>
> I know this is a very simple question ( I apologize in advanced ) but
> I am just curious what the accepted approach is for passing a
> variable sized array from the kernel to user-space.
>
> I am familiar with sending data from user to kernel, but the other way
> seems a bit more complicated.
>
> Should I create two ioctls i.e. one for getting the size, and then
> another for sending it down to the driver? Or, perhaps create a MAX
> static sized buffer and have the driver populate that?
>
> Thanks,
> J
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer
>
>
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer
>
—
NTDEV is sponsored by OSR
For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars
To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer
