getNumberOfSectors() returns incorrect number

[Magnus Karlsson]

The function getNumberOfSectors is not returning the actual sector count. For example, on one of my 32 GB sd-cards with 64845824 sectors the function returns 64838340. This means that the whole disk is not read when doing an image read from the SD card.

Below is a version of getNumberOfSectors that returns the correct number:

unsigned long long getNumberOfSectors(HANDLE handle, unsigned long long *sectorsize)
{
 DWORD junk;
 DISK_GEOMETRY geometry;
 PARTITION_INFORMATION part;
 BOOL bResult;
 bResult = DeviceIoControl(handle, IOCTL_DISK_GET_PARTITION_INFO, NULL, 0, &part, sizeof(part), &junk, NULL) &
   DeviceIoControl(handle, IOCTL_DISK_GET_DRIVE_GEOMETRY, NULL, 0, &geometry, sizeof(geometry), &junk, NULL);
 if (!bResult)
 {
  char *errormessage=NULL;
  FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_ALLOCATE_BUFFER, NULL, GetLastError(), 0, errormessage, 0, NULL);
  QMessageBox::critical(NULL, "Device Error", QString("An error occurred when attempting to get the device's geometry.\nError %1: %2").arg(GetLastError()).arg(errormessage));
  LocalFree(errormessage);
  return 0;
 }
 if (sectorsize != NULL)
  *sectorsize = (unsigned long long)geometry.BytesPerSector;
 return (unsigned long long)part.PartitionLength.QuadPart / (unsigned long long)geometry.BytesPerSector;
}

[Jeff B (skydiver38)]

Thanks for the patch. So the length of a partition (in sectors) is not necessarily equal to the cylinders * tracks * sectors given in the DISK_GEOMETRY struct. Good to know, I guess.

I wonder if using the "disk size (in bytes)" field of the DISK_GEOMETRY_EX structure would report the same as C*T*S*Bytes (per sector), or the same as the partition info.

At any rate - thanks again for the patch, and I'll get it incorporated for a future release.

[Magnus Karlsson (u-magnds-b)]

I changed my code to use the DISK_GEOMETRY_EX like you suggest and this seems to work fine too.

The is the patch based on this verson:

unsigned long long getNumberOfSectors(HANDLE handle, unsigned long long *sectorsize)
{
 DWORD junk;
 DISK_GEOMETRY_EX diskgeometry;
 BOOL bResult;

 bResult = DeviceIoControl(handle, IOCTL_DISK_GET_DRIVE_GEOMETRY_EX, NULL, 0, &diskgeometry, sizeof(diskgeometry), &junk, NULL);
 if (!bResult)
 {
  char *errormessage=NULL;
  FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_ALLOCATE_BUFFER, NULL, GetLastError(), 0, errormessage, 0, NULL);
  QMessageBox::critical(NULL, "Device Error", QString("An error occurred when attempting to get the device's geometry.\nError %1: %2").arg(GetLastError()).arg(errormessage));
  LocalFree(errormessage);
  return 0;
 }

 if (sectorsize != NULL)
  *sectorsize = (unsigned long long)diskgeometry.Geometry.BytesPerSector;
 return (unsigned long long)diskgeometry.DiskSize.QuadPart / (unsigned long long)diskgeometry.Geometry.BytesPerSector;

}

[Tobin Davis (gruemaster)]

Gah! I had already incorporated the first patch and was testing it. Let me get the new patch incorporated. I plan on pushing the new release today (while I have a few spare cycles).

[Tobin Davis (gruemaster)]

On Wed, 2012-02-15 at 21:36 +0000, Jeff B wrote:
> Thanks for the patch. So the length of a partition (in sectors) is not
> necessarily equal to the cylinders * tracks * sectors given in the
> DISK_GEOMETRY struct. Good to know, I guess.
>
> I wonder if using the "disk size (in bytes)" field of the
> DISK_GEOMETRY_EX structure would report the same as C*T*S*Bytes (per
> sector), or the same as the partition info.
>
> At any rate - thanks again for the patch, and I'll get it incorporated
> for a future release.
>

I'm already incorporating that patch (and now the revised one). I also
(finally) have your md5 patch in the tree. I'll push it up soon, then
prep a release today.

I have been over-swamped with work since returning from our rally in
Budapest. This is the first time I have had a chance to get back to the
code.

--
Tobin Davis

If you teach your children to like computers and to know how to gamble
then they'll always be interested in something and won't come to no real harm.

[Tobin Davis (gruemaster)]

In the future, could you add changes as a patch diff? I'm sure there are tools in Windows to do this. If not, use Bazaar Explorer for WIndows (http://bazaar.canonical.com).

[Magnus Karlsson (u-magnds-b)]

One more comment:
The cylinders, tracks/cylinder, sectors/track data is something that Windows makes up since the sd-card or thumbdrive is not structured like that, they only report the total size in blocks, where 1 block = 512 bytes.

Since the c, t, s values are integers, Windows must pick c, t and s values so that the formula c * t * s gives a conservative approximation of the disk size, but this is very likely not the correct size.

[Jeff B (skydiver38)]

Magnus - great to know. Thanks.

Tobin - OK. Good luck with the release, and I'll just take your update for this.

Is this bug really a duplicate of Bug 786060? (https://bugs.launchpad.net/win32-image-writer/+bug/786060) It talks about modifying the sector calculation as well.

[Tobin Davis (gruemaster)]

Ah, yes. Will tag that bug as fixed as well. Thanks for remembering this one.

[Dennis Jackson (dennis-m-jackson)]

Greetings -

     Just tried v0.5 of Win32 Disk Imager. Executable errors out at startup
indicating that libgcc_s_dw2-1.dll is missing. I located and downloaded
that DLL and installed it in the same directory as Win32 Disk Imager.exe.
Errors out again, but missing DLL is now reported to be libstdc++-6.dll. I
notice that another user has already reported this error, report number is
187903. Thanks.

----- Original Message -----
From: "Tobin Davis" <email address hidden>
To: <email address hidden>
Sent: Wednesday, February 15, 2012 5:43 PM
Subject: [Bug 932321] Re: getNumberOfSectors() returns incorrect number

Ah, yes. Will tag that bug as fixed as well. Thanks for remembering
this one.

--
You received this bug notification because you are subscribed to a
duplicate bug report (912259).
https://bugs.launchpad.net/bugs/932321

Title:
  getNumberOfSectors() returns incorrect number

Status in Image Writer for Windows:
  New

Bug description:
  The function getNumberOfSectors is not returning the actual sector
  count. For example, on one of my 32 GB sd-cards with 64845824 sectors
  the function returns 64838340. This means that the whole disk is not
  read when doing an image read from the SD card.

  Below is a version of getNumberOfSectors that returns the correct
  number:

  unsigned long long getNumberOfSectors(HANDLE handle, unsigned long long
*sectorsize)
  {
  DWORD junk;
  DISK_GEOMETRY geometry;
  PARTITION_INFORMATION part;
  BOOL bResult;
  bResult = DeviceIoControl(handle, IOCTL_DISK_GET_PARTITION_INFO, NULL, 0,
&part, sizeof(part), &junk, NULL) &
  DeviceIoControl(handle, IOCTL_DISK_GET_DRIVE_GEOMETRY, NULL, 0, &geometry,
sizeof(geometry), &junk, NULL);
  if (!bResult)
  {
  char *errormessage=NULL;
  FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_ALLOCATE_BUFFER,
NULL, GetLastError(), 0, errormessage, 0, NULL);
  QMessageBox::critical(NULL, "Device Error", QString("An error occurred
when attempting to get the device's geometry.\nError %1:
%2").arg(GetLastError()).arg(errormessage));
  LocalFree(errormessage);
  return 0;
  }
  if (sectorsize != NULL)
  *sectorsize = (unsigned long long)geometry.BytesPerSector;
  return (unsigned long long)part.PartitionLength.QuadPart / (unsigned long
long)geometry.BytesPerSector;
  }

To manage notifications about this bug go to:
https://bugs.launchpad.net/win32-image-writer/+bug/932321/+subscriptions