Re: [JANITOR PROPOSAL] Switch ioctl functions to ->unlocked_ioctl

!MAILaRCHIVE_VOTE_RePLACE
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: Alasdair G Kergon <agk@...>
Cc: Matt Mackall <mpm@...>, Paolo Ciarrocchi <paolo.ciarrocchi@...>, Andi Kleen <andi@...>, <linux-kernel@...>, <kernel-janitors@...>, <gorcunov@...>
Date: Thursday, January 10, 2008 - 6:55 pm

On Thursday 10 January 2008 03:39, Alasdair G Kergon wrote:

Nice.  This removes a deadlock we hit, where if creating a device mapper 
target blocks indefinitely (say on network IO) then nobody else can 
complete a device mapper operation because BKL is held.  If completing 
the device create depends on some other device mapper operation, then 
it is game over.

Our current workaround is to test for and drop BKL, ugh.  Thanks for the 
cleanup.

Regards,

Daniel
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
Re: [JANITOR PROPOSAL] Switch ioctl functions to -&gt;unlock..., Alasdair G Kergon, (Thu Jan 10, 7:39 am)
Re: [JANITOR PROPOSAL] Switch ioctl functions to ->unlock..., Daniel Phillips, (Thu Jan 10, 6:55 pm)
Re: [JANITOR PROPOSAL] Switch ioctl functions to -&gt;unlock..., Christoph Hellwig, (Thu Jan 10, 4:34 am)