From: Tvrtko Ursulin <tvrtko.ursu...@intel.com>
I spotted a few small issues in the recent added SGL code so am sending some
patches to tidy this.
My motivation was looking at sgl_alloc_order to potentially use from the i915
driver, with a small addition to support fall-back to smaller order allocation
if so was requested.
But then I realized scatterlist table allocated with sgl_alloc_order is not
compatible with being put into the sg_table container due different allocator
If it had used the standard chained scatterlist allocation, it would benefit
from less memory pressure for small-order large length allocations (in other
words large nent - as it stands it can request really large kmalloc allocations
for those cases), but to convert it to use that looks to would require some
refactoring of the SGL API users. So I wasn't sure how feasible would that be.
There were some other unclear bits to me in the SGL API, like why so much API
some of which is unused, so I tried to trim that as well.
So don't know - comments are welcome.
Tvrtko Ursulin (6):
lib/scatterlist: Tidy types and fix overflow checking in
lib/scatterlist: Skip requesting zeroed allocations in sgl_alloc_order
lib/scatterlist: Do not leak pages when high-order allocation fails
lib/scatterlist: Unexport some trivial wrappers
lib/scatterlist: Drop unused sgl_free_order
lib/scatterlist: Drop order argument from sgl_free_n_order
drivers/target/target_core_transport.c | 2 +-
include/linux/scatterlist.h | 36 +++++++++++---
lib/scatterlist.c | 91 +++++++++++-----------------------
3 files changed, 57 insertions(+), 72 deletions(-)
Cc: Bart Van Assche <bart.vanass...@wdc.com>
Cc: Hannes Reinecke <h...@suse.com>
Cc: Johannes Thumshirn <jthumsh...@suse.de>
Cc: Jens Axboe <ax...@kernel.dk>