[Openib-windows] RE: MTHCA: FW update paths

[Leonid Keller]

Mostly applied.
See comment for revision 272

> -----Original Message-----
> From: Fab Tillier [mailto:ftillier at silverstorm.com] 
> Sent: Saturday, April 01, 2006 12:30 AM
> To: Leonid Keller
> Cc: openib-windows at openib.org
> Subject: MTHCA: FW update paths
> 
> Hi Leo,
> 
> Now that we have safe CR space mappings into user space, I 
> have a couple of questions.
> 
> Should we eliminate the CR space access through configuration 
> cycles?  There's an issue in that code wherein the PCI bus 
> interface reference is not properly released if the user-mode 
> application exits abnormally.
> 
> Should a single application be allowed to map the CR space 
> multiple times?  The code currently does, but I don't know if 
> it makes any sense.  What would be a usage mode of doing so?
> 
> Should multiple applications be allowed to map the CR space 
> simultaneously?
> This opens up the possibility for multiple users (in 
> different remote desktop
> sessions) to attempt to update the firmware at the same time, 
> which seems like a flaw.
> 
> I'll code up patches for these if you think it makes sense to 
> change them.
> 
> Lastly, (and I know you're not the maintainer), the flint 
> usage output isn't very helpful - I had to go look at the 
> code to find out how to open a device.
> The usage output seems targeted at Linux (based on the device 
> names).  The output of the query command also is not so 
> friendly.  Could the query output be changed to match that of 
> the hcafwupdate tool when used with the -j option?
> Right now the flint tool seems like a very very low level 
> tool, and I fear that it could result in issues with failed 
> firmware updates, not to mention frustration in how to run 
> it.  Also, a way to correlate the device names used to the 
> devices themselves would be great.
> 
> Thanks,
> 
> - Fab
> 
>