This patch moves memset() calls from p80211netdev_ethtool()
after copy_from_user() call in order to avoid unnecessary
instruction in case copy_from_user() fails.
Signed-off-by: Claudiu Beznea <[email protected]>
---
drivers/staging/wlan-ng/p80211netdev.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/staging/wlan-ng/p80211netdev.c b/drivers/staging/wlan-ng/p80211netdev.c
index 88255ce..3723b8c 100644
--- a/drivers/staging/wlan-ng/p80211netdev.c
+++ b/drivers/staging/wlan-ng/p80211netdev.c
@@ -465,12 +465,12 @@ static int p80211netdev_ethtool(wlandevice_t *wlandev, void __user *useraddr)
struct ethtool_drvinfo info;
struct ethtool_value edata;
- memset(&info, 0, sizeof(info));
- memset(&edata, 0, sizeof(edata));
-
if (copy_from_user(ðcmd, useraddr, sizeof(ethcmd)))
return -EFAULT;
+ memset(&info, 0, sizeof(info));
+ memset(&edata, 0, sizeof(edata));
+
switch (ethcmd) {
case ETHTOOL_GDRVINFO:
info.cmd = ethcmd;
--
1.9.1
On Tue, Mar 15, 2016 at 09:53:15PM +0200, Claudiu Beznea wrote:
> This patch moves memset() calls from p80211netdev_ethtool()
> after copy_from_user() call in order to avoid unnecessary
> instruction in case copy_from_user() fails.
Does that really matter? If an error happens, then we are on a "slow
path" and the extra memset didn't matter at all. Don't move code around
for no good reason, and if you think it makes things faster, then you
have to be able to prove it somehow :)
thanks,
greg k-h
On Tue, Mar 22, 2016 at 03:02:00PM +0200, Claudiu Beznea wrote:
> I was guided by the principle that every instruction counts, no matter the
> execution context.
But you have to be able to actually measure the difference, and here you
can't.
thanks,
greg k-h