On Sun, 30 Oct 2005 17:47:58 -0500, Alan Cox <alan(a)redhat.com> wrote:
I'd oppose changing. usb-storage finally works [...]
This is exactly what I told myself for 3 long years.
The ub consists of pure and distilled anger over oopses and lockups
in usb-storage, which overflowed the cup.
Essentially I wrote it when we betaed RHEL 4 with completely busted
SCSI stack. Remember the "Illegal transition" oopses? I thought, why
do I have to put up with this crap? Why did I do it for so long?!
And the answer was, "because I did not want to rock the boat, because
I wanted to play nicely with others", and not for any technical reason.
I saw too many people who pushed their own code because they were too
dumb to understand existing code, and I did not want to be like them.
It took me the horrible disaster of 2.4.9 to realize my folly.
That's how ub started. Then there was a period of calm, with some
small problems like interaction with VM. No problem, I thought, just
do not use ext3 with usb-storage and you're golden. So, ub went
nowhere. Users complained of regressions and poorly made mechanism
of sharing of devices (it required a kernel rebuild).
Then came the 2.6.13. It's a release where you pull the USB key, you
get an oops in khubd, no ifs or buts. It's as if nobody tests anything
anymore. So I wrote libusual, which allows for perfect sharing.
Now, if usb-storage works well for you, you can do this:
echo 'options libusual bias="usb-storage"' >> /etc/modprobe.conf
You'll never know that anything has changed.
But honestly, I do not feel strongly about all this, not until someone
assigns me next bug in usb-storage. It works, for now, and I do feel
like arguing with you. Just wanted to let you know that I went through
the same denial already and hoped that I would save you some
P.S. Another e-mail from yesterday: