[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#279696: ub driver:Issues with makedev, improper major number and slow performance



tag +pending 279696
quit

Hi Gene,

I took a closer look into this and in particular Bug#278237.
I agree that the driver should be disabled for now,
at at the very least it seems green and problematic.
I will make the change in SVN and it should appear in the next 
release of 2.6.9, whenever that is.

As an asside, 2.6.8 is the kernel that will go into sarge,
unless some calamity occurs. 2.6.9 is out there for people to
play with at this stage.

-- 
Horms

On Fri, Nov 05, 2004 at 03:22:05PM +0900, Horms wrote:
> On Thu, Nov 04, 2004 at 09:16:18PM -0500, Gene Cumm wrote:
> > There is a bug on makedev (#278237) concerning the lack of support for 
> > the new and experimental ub driver.  This driver was prematurely 
> > released using an experimental major number (125/b) and about 9 days 
> > later was assigned a major number (180/b).  Until makedev is updated, 
> > the device files must be created manually (not too big of a deal, if you 
> > realize the problem and the solution).
> 
> That should be easy enough to resolve, I will look into it.
> 
> > More importantly, this is a low performance driver designed for low 
> > performance flash drives however many USB/ATA chipsets also use this 
> > same identification: Class MassStorage(0x08) Subclass SCSI(0x06) 
> > Protocol Bulk(0x50).  I've read nothing but complaints about this 
> > driver, the issues it causes and the performance it lacks.  One 
> > individual reported 67% loss in performance.  I know that I have 
> > personally seen a drop in performance once I could use the driver.
> > 
> > I was about to submit a bug on one of the binary packages until I saw 
> > this one tonight.  Until this issue is resolved, 2.6.8 looks a lot better.
> 
> I am confused. Does it cause the system to slow down when
> you are not using this driver? Is there an alternate or
> older version of the driver that is faster? In short, does
> including ub in the build take away any feature that
> would be present if it was not built?



Reply to: