drm: Use a little stash on the stack to avoid kmalloc in most DRM ioctls.
The kmalloc was taking up about 1.5% of the CPU on an ioctl-heavy workload
(x11perf -aa10text on 965). Initial results look like they have a
corresponding improvement in performance for aa10text, but more numbers might
not hurt.
Thanks to ajax for pointing out this performance regression I'd introduced
back in 2007.
[airlied: well I introduced it sneakily inside Eric's patch]
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
index c26ee08..c4ada8b 100644
--- a/drivers/gpu/drm/drm_drv.c
+++ b/drivers/gpu/drm/drm_drv.c
@@ -421,6 +421,7 @@
drm_ioctl_t *func;
unsigned int nr = DRM_IOCTL_NR(cmd);
int retcode = -EINVAL;
+ char stack_kdata[128];
char *kdata = NULL;
atomic_inc(&dev->ioctl_count);
@@ -459,10 +460,14 @@
retcode = -EACCES;
} else {
if (cmd & (IOC_IN | IOC_OUT)) {
- kdata = kmalloc(_IOC_SIZE(cmd), GFP_KERNEL);
- if (!kdata) {
- retcode = -ENOMEM;
- goto err_i1;
+ if (_IOC_SIZE(cmd) <= sizeof(stack_kdata)) {
+ kdata = stack_kdata;
+ } else {
+ kdata = kmalloc(_IOC_SIZE(cmd), GFP_KERNEL);
+ if (!kdata) {
+ retcode = -ENOMEM;
+ goto err_i1;
+ }
}
}
@@ -483,7 +488,7 @@
}
err_i1:
- if (kdata)
+ if (kdata != stack_kdata)
kfree(kdata);
atomic_dec(&dev->ioctl_count);
if (retcode)