Aleš Nesrsta wrote:
> Hi Vladimir,
>
> I have some news - I made extensive testing and researching and I
> discovered some things related to OHCI and generally to USB MS or USB
> itself.
>
> I tried to fix them and it looks like I was probably successful in some
> cases:
>
> - multiple LUNs related bug (meaning of value from device & coding into
> CBW)
> - bulk transport toggling related bug - on devices with endpoints with
> the same "low" nuber (e.g. 01 & 81)
> - bug related to devices with max. packet size for control pipe 0 lower
> than 64
> - problem with one kind of OHCI BIOS legacy support maybe solved
> - another bug - some devices hangs because of Bulk-only Mass Storage
> Reset used during opening when it is not expected by device
> - my OHCI done-head loop should not hang anymore - timeout was added
> - "ls" should not hang anymore on USB MS device with not ready unit -
> timeout was added in SCSI open unit ready loop
>   
I hope in future we'll be able to do useful operations during wait.
> - experimentally added support for another subclasses of USB MS devices
> (as all are SCSI based - it possibly should work, at least it is working
> with one my device with subclass SFF-8070i - but I don't have another
> devices to test)
>
> In fact, all my USB MS devices are working now with grub2 with exception
> of one camera which have not Bulk-Only but CBI protocol.
>
> What I was not solved till now:
> - Some devices are not working on UHCI
> - If I connect some device and load uhci module, I cannot connect
> another device without rebooting. On OHCI it is possible without
> rebooting in this sequence:
> 	rmmod usbms
> 	rmmod ohci
> 	(unplug old device and plug new one)
> 	insmod ohci
> 	insmod usbms
>   
In my yeeloongfw branch I have hotplug support for ATA. I'm planning on
adding usbms hotplugging support but first I want to make Geode USB work.
> - It looks like my OHCI is not properly reporting STALL state of pipe -
> it is probably reason why done-head loop hanged if some problem happened
> in communication.
>   
In general there should be timeout in any loop which may hang if device
doesn't behave the way it's supposed to.

The fixes I've come up with are available in yeeloongfw branch. I've
adjusted your previous patch to work on top of yeeloongfw branch.
> I want to prepare patch and send it here but I cannot get actual bzr
> code. I would want have it at least because grub_memalign problem is not
> solved in last grub version wich I have (1.98) and I have to solve it by
> some workarounds (maybe it is also reason why UHCI does not work
> properly on my computer, I did not any workaround in its code yet).
>
> I cannot get actual bzr code because of this error:
> $ bzr get http://bzr.savannah.gnu.org/r/grub
> bzr: ERROR: Not a branch:
> "http://bzr.savannah.gnu.org/r/grub/.bzr/branch/".
>
> I tried to found some advice in man bzr and forums but I was not
> successful. It looks like problem is that there is really not existing
> folder ".../r/.bzr/branch" (I found "branch-lock" and "repository" only)
> - should I use some another address or some option ?
> Do You have some experiences with such problem ?
> Could You put me in right direction...?
>
> (I tried to download CVS source code - it works - but it looks outdated.
> I also tried register ssh public code and made access via ssh to
> sftp://starous@bzr.savannah.gnu.org/srv/bzr/grub - but I cannot login,
> there is ssh debug message that it is not possible to find
> authentication method...)
>
> Best regards
> Ales
>
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/grub-devel
>
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko