dmaengine: provide a common 'issue_pending_all' implementation
authorDan Williams <dan.j.williams@intel.com>
Tue, 6 Jan 2009 18:38:14 +0000 (11:38 -0700)
committerDan Williams <dan.j.williams@intel.com>
Tue, 6 Jan 2009 18:38:14 +0000 (11:38 -0700)
commit2ba05622b8b143b0c95968ba59bddfbd6d2f2559
treeb7b72d02a993ff2ba731d6608f4ab8ce87482bcb
parentbec085134e446577a983f17f57d642a88d1af53b
dmaengine: provide a common 'issue_pending_all' implementation

async_tx and net_dma each have open-coded versions of issue_pending_all,
so provide a common routine in dmaengine.

The implementation needs to walk the global device list, so implement
rcu to allow dma_issue_pending_all to run lockless.  Clients protect
themselves from channel removal events by holding a dmaengine reference.

Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
crypto/async_tx/async_tx.c
drivers/dma/dmaengine.c
include/linux/async_tx.h
include/linux/dmaengine.h
net/core/dev.c