From: Alexander Usyskin <alexander.usys...@intel.com> The read callbacks for the fixed address clients, that don't have flow control are built now on the receive path. In order to have a single allocation place we remove the allocation from the read request.
Signed-off-by: Alexander Usyskin <alexander.usys...@intel.com> Signed-off-by: Tomas Winkler <tomas.wink...@intel.com> --- V2: resend drivers/misc/mei/client.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/drivers/misc/mei/client.c b/drivers/misc/mei/client.c index 2a09db86e50e..05cacc3e718f 100644 --- a/drivers/misc/mei/client.c +++ b/drivers/misc/mei/client.c @@ -1459,17 +1459,15 @@ int mei_cl_read_start(struct mei_cl *cl, size_t length, const struct file *fp) return -ENOTTY; } + if (mei_cl_is_fixed_address(cl)) + return 0; + /* always allocate at least client max message */ length = max_t(size_t, length, mei_cl_mtu(cl)); cb = mei_cl_alloc_cb(cl, length, MEI_FOP_READ, fp); if (!cb) return -ENOMEM; - if (mei_cl_is_fixed_address(cl)) { - list_add_tail(&cb->list, &cl->rd_pending); - return 0; - } - rets = pm_runtime_get(dev->dev); if (rets < 0 && rets != -EINPROGRESS) { pm_runtime_put_noidle(dev->dev); -- 2.7.4