linux/mm/dmapool.c, branch v3.2.38

mm: dmapool: use provided gfp flags for all dma_alloc_coherent() calls

2013-01-03T03:33:36Z

commit 387870f2d6d679746020fa8e25ef786ff338dc98 upstream. dmapool always calls dma_alloc_coherent() with GFP_ATOMIC flag, regardless the flags provided by the caller. This causes excessive pruning of emergency memory pools without any good reason. Additionaly, on ARM architecture any driver which is using dmapools will sooner or later trigger the following error: "ERROR: 256 KiB atomic DMA coherent pool is too small! Please increase it with coherent_pool= kernel parameter!". Increasing the coherent pool size usually doesn't help much and only delays such error, because all GFP_ATOMIC DMA allocations are always served from the special, very limited memory pool. This patch changes the dmapool code to correctly use gfp flags provided by the dmapool caller. Reported-by: Soeren Moch Reported-by: Thomas Petazzoni Signed-off-by: Marek Szyprowski Tested-by: Andrew Lunn Tested-by: Soeren Moch Signed-off-by: Ben Hutchings

mm: fix implicit stat.h usage in dmapool.c

2011-10-31T13:20:12Z

The removal of the implicitly everywhere module.h and its child includes will reveal this implicit stat.h usage: mm/dmapool.c:108: error: ‘S_IRUGO’ undeclared here (not in a function) Signed-off-by: Paul Gortmaker

mm: Map most files to use export.h instead of module.h

2011-10-31T13:20:12Z

The files changed within are only using the EXPORT_SYMBOL macro variants. They are not using core modular infrastructure and hence don't need module.h but only the export.h header. Signed-off-by: Paul Gortmaker

devres: fix possible use after free

2011-07-26T03:57:14Z

devres uses the pointer value as key after it's freed, which is safe but triggers spurious use-after-free warnings on some static analysis tools. Rearrange code to avoid such warnings. Signed-off-by: Maxin B. John Reviewed-by: Rolf Eike Beer Acked-by: Tejun Heo Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm/dmapool.c: use TASK_UNINTERRUPTIBLE in dma_pool_alloc()

2011-01-14T01:32:48Z

As it stands this code will degenerate into a busy-wait if the calling task has signal_pending(). Cc: Rolf Eike Beer Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm/dmapool.c: take lock only once in dma_pool_free()

2011-01-14T01:32:48Z

dma_pool_free() scans for the page to free in the pool list holding the pool lock. Then it releases the lock basically to acquire it immediately again. Modify the code to only take the lock once. This will do some additional loops and computations with the lock held in if memory debugging is activated. If it is not activated the only new operations with this lock is one if and one substraction. Signed-off-by: Rolf Eike Beer Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: add a might_sleep_if() to dma_pool_alloc()

2010-10-26T23:52:08Z

Buggy drivers (e.g. fsl_udc) could call dma_pool_alloc from atomic context with GFP_KERNEL. In most instances, the first pool_alloc_page call would succeed and the sleeping functions would never be called. This allowed the buggy drivers to slip through the cracks. Add a might_sleep_if() checking for __GFP_WAIT in flags. Signed-off-by: Dima Zavin Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

dmapools: protect page_list walk in show_pools()

2009-07-01T01:56:00Z

show_pools() walks the page_list of a pool w/o protection against the list modifications in alloc/free. Take pool->lock to avoid stomping into nirvana. Signed-off-by: Thomas Gleixner Signed-off-by: Matthew Wilcox Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

dmapool: enable debugging for CONFIG_SLUB_DEBUG_ON too

2008-04-28T15:58:20Z

Previously it was only enabled for CONFIG_DEBUG_SLAB. Not hooked into the slub runtime debug configuration, so you currently only get it with CONFIG_SLUB_DEBUG_ON, not plain CONFIG_SLUB_DEBUG Acked-by: Matthew Wilcox Signed-off-by: Andi Kleen Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

pool: Improve memory usage for devices which can't cross boundaries

2007-12-04T15:39:58Z

The previous implementation simply refused to allocate more than a boundary's worth of data from an entire page. Some users didn't know this, so specified things like SMP_CACHE_BYTES, not realising the horrible waste of memory that this was. It's fairly easy to correct this problem, just by ensuring we don't cross a boundary within a page. This even helps drivers like EHCI (which can't cross a 4k boundary) on machines with larger page sizes. Signed-off-by: Matthew Wilcox Acked-by: David S. Miller