Re: Win32 FILE_BASIC_INFORMATION.*Time.QuadPart - which value should be used if filesystem does not support it, so Cygwin /usr/bin/stat lists the value as '-'?

[Corinna Vinschen <corinna-cygwin@cygwin.com>]

On Apr  9 22:30, Brian Inglis via Cygwin wrote:
> On 2024-04-09 15:14, Corinna Vinschen via Cygwin wrote:
> > On Apr  5 04:26, Martin Wege via Cygwin wrote:
> > > On Fri, Apr 5, 2024 at 2:05 AM Martin Wege <martin.l.wege@gmail.com> wrote:
> > > > I have problems with debugging, so a quick help would be appreciated,
> > > > as I cannot figure this out after several hours of digging.
> > > > 
> > > > Cygwin /usr/bin/stat returns "Birth: -" for some files. Which value
> > > > must the CreationTime member of FILE_BASIC_INFORMATION have to cause
> > > > /usr/bin/stat ti return "-"? 0, -1, or something else?
> > > 
> > > In a related matter:
> > > The Win32 FILE_BASIC_INFORMATION structure defines four time values:
> > > 
> > > LARGE_INTEGER CreationTime;
> > > LARGE_INTEGER LastAccessTime;
> > > LARGE_INTEGER LastWriteTime;
> > > LARGE_INTEGER ChangeTime;
> > > 
> > > How can a filesystem indicate if it does not support a particular
> > > timestamp, such as ChangeTime? Should ChangeTime.QuadPart then be -1,
> > > -2 or 0, or another value?
> > 
> > I'm not aware of a filesystem not supporting ChangeTime, that is,
> > st_ctime.  Usually only CreationTime (st_birthtime) is missing.
> 
> R/O media like CD/DVD-R or FS w/o write support?

Oh yes, that makes sense, CDFS and the likes of them.

> > I think setting the timestamp to 0 works for indicating that this kind
> > of timestamp is not supported.  Cygwin is handling Windows timestamps
> > this way, but I can't find this in documentation ATM.
> 
> See upthread?:
> 
> Caller or application can set 0 to mean keep/return current value, caller or
> driver can set -1 to mean don't update/return current value:
> 
> https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/wdm/ns-wdm-_file_basic_information#remarks

I'm aware of that.  This remarks session is copy/pasted in other MSFT
articles, too.  But it doesn't really answer Martin's question.

It describes what a consumer (application or driver) is supposed to set
the timestamps to when passing timestamps to a filesystem driver via
ZwSetInformationFile (or its friends on the lower OS levels).

It does *not* describe the other direction of the call stack, i. e.,
what a filesystem driver is supposed to set a timestamp to, if its
underlying filesystem doesn't support this kind of timestamp.

And that's the direction Martin is asking about.


Corinna