Re: "legacy" usb stack fixes (was: Re: HEADSUP new usb code coming in.)

On 08/24/08 06:57, Antony Mawer wrote:
> On 23/08/2008 8:08 PM, Volker wrote:
>> On 12/23/-58 20:59, Antony Mawer wrote:
>>> M. Warner Losh wrote:
>>>> In message: <48AB566B.5010507_at_mawer.org>
>>>>             Antony Mawer <fbsd-current_at_mawer.org> writes:
>>>> : Warner Losh wrote:
>>>> : > From: Bakul Shah <bakul_at_bitblocks.com>
>>>> : > Subject: Re: HEADSUP new usb code coming in. : > Date: Tue, 19 Aug
>>>> 2008 14:18:13 -0700
>>>> : > : >> On Tue, 19 Aug 2008 22:44:20 +0200 Hans Petter Selasky
>>>> <hselasky_at_c2i.net>  wrote:
>>>> : >>> New stuff (all of which I can remember right now):
>>>> : >>     ...
>>>> : >>
>>>> : >> Accidentally unplugging a mounted USB disk (without
>>>> : >> unmounting it) resulted in a hang or a crash.  Is this fixed?
>>>> : > : > That's fixed in -current right now with the old stack.  It
>>>> isn't a usb
>>>> : > issue at all, but a buffer cache issue.
>>>> : : Is this change that is likely to be MFC'd in time for 7.1? And/or
>>>> is : there a specific patch that can manually be applied to -STABLE to
>>>> fix this?
>>>>
>>>> I should spend the time to dig into the changes in current.  There
>>>> turned out to be several little changes...  And I need to verify all
>>>> the edge cases were covered...
>>> I'd be happy to test patches if you do end up doing this.. it would be
>>> really nice to have in 7.1, or at least available as a patchset if it
>>> isn't suitable for MFC (eg. ABI changes)...
>>
>> I'm a bit behind with reading emails. Please forgive me if this has
>> already been answered.
>>
>> Don't expect the new USB stack for 7.1-R. It's too short and the new USB
>> stack will introduce an ABI breakage. For that, all drivers written for
>> the old USB stack need to be rewritten and I guess, we need to take care
>> about 3rd party developers and inform them in advance about that massive
>> change. I would not wonder if this will never get MFC'd but I don't know
>> actually.
> 
> This wasn't about the new USB stack -- we were discussing the buffer
> cache and CAM-related fixes that prevents the system from panic'ing when
> a USB device is unplugged without first unmounting the filesystem. These
> patches are in HEAD with the existing USB stack. :-)
> 
> --Antony
> 

Antony,

please forgive me. While re-reading the thread partially, I've also seen
your question was about the old usb stack.

That does happen if $SUBJECT does not match $CONTENT (you could have
changed $SUBJECT when driving the discussion into another direction).

Anyway, I've already had those crashes even with the "new" usb stack
(but it doesn't happen everytime - YMMV).

Volker