Umm... let's start with the fact that 2.5.x is polluted with all sorts of
other changes, none of which apply to 2.4.x

I've asked for someone to help with concurrent 2.4.x and 2.5.x development,
but nobody seems interested.  I only have two hands and one brain, so....

Heck, at one point I even asked if you were interested in porting some
changes from 2.4.x to 2.5.x, and you realized it would be too major.  I
don't think that testing these in a 2.5.x environment would really tell us
anything about a 2.4.x environment.

The reason I dropped you this patch is that the next few, which radically
fix the error handling (abort, reset, etc) logic depends on this.  I know
smaller patches are preferred over larger ones.

And, the truth is, my original description was somewhat lacking.  This
actually does fix some bugs -- for example, I can no longer crash my system
when a device reset or command abort occurs.  It's a rare condition, but
the fix is notable.  Resets can no longer take forever (timeout
implemented), endpoints are always left in a sane state, and the return
codes are now consistent.

Prior to this, a command being aborted would crash a system 30-60% of the
time.  It was very hit-or-miss, depending on lots of other factors.  Now,
unless the HCD crashes, we're pretty much guaranteed to at least keep the
system up, tho the device may be unusable.

The next patches, which are notably more major, _really_ fix the
abort/reset paths, and eliminate (we think) all the possible race
conditions, including SMP problems.

Oh, and these were originally prototypes by Alan Stern, not Rowland Stern.
Lord knows where my brain was on that one.  The code was originally written
back in Jan/Feb, and has been under review and test since...

Perhaps we should just circulate this among the developers for a while, and
then send it along?  Tho it's been circulating on the usb-storage
developers for a while already....

Matt

On Sun, Mar 31, 2002 at 04:05:47PM -0800, Greg KH wrote:
> On Sat, Mar 30, 2002 at 07:38:09PM -0800, Matthew Dharm wrote:
> > Attached is a patch to usb-storage against 2.4.19-pre5.  Greg, please send
> > to Marcello for inclusion.
> > 
> > This patch consolidates a great deal of common code in the URB submission
> > paths, and in the reset paths.  This is in preparation for the new
> > error-handling state machine.
> > 
> > These changes are courtsey of Rowland Stern.  I've tested them, and they
> > look good.  More changes will follow shortly.
> 
> Um no, I will not apply this.  This is new work that doesn't look like
> it fixes any bug, and has not appeared in the 2.5 tree yet for testing.
> 
> Sorry if this is a shock to some people, but new work needs to be done
> in the 2.5 tree, and then it can be backported to 2.4.  I know I haven't
> done a great job in following this rule, but I'm going to try to be
> better at it in the future.
> 
> Is there any reason this patch can't be applied to 2.5 instead, and then
> to 2.4 after it's proven itself?
> 
> thanks,
> 
> greg k-h
> 
> _______________________________________________
> [EMAIL PROTECTED]
> To unsubscribe, use the last form field at:
> https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

-- 
Matthew Dharm                              Home: [EMAIL PROTECTED] 
Maintainer, Linux USB Mass Storage Driver

You suck Stef.
                                        -- Greg 
User Friendly, 11/29/97

Attachment: msg05436/pgp00000.pgp
Description: PGP signature

Reply via email to