RE: [PATCH V3 3/7] Drivers: hv: vmbus: add APIs to send/recv hvsock packet and get the r/w-ability

From: Dan Carpenter
Sent: Thursday, July 23, 2015 18:25
On Thu, Jul 23, 2015 at 01:10:57PM +0300, Dan Carpenter wrote:

quoted

In this specific case, writing it as "if (ret != 0)" caused the bug.  If
we had written it as "if (ret) return ret;" then there are no zeroes so
wouldn't have been any temptation to return the zero instead of the ret.

I did a search to see if returning the zero instead of the ret was a
common mistake and it seems like it might be.  I did:

grep 'if (ret != 0)' drivers/   -r -A1 -n | grep "return 0;" | perl -ne 's/.c-(\d+)-/.c:$1/;
print'

drivers/gpu/drm/nouveau/nouveau_display.c:111                   return 0;
drivers/regulator/wm8400-regulator.c:47         return 0;
drivers/platform/x86/dell-laptop.c:859          return 0;
drivers/media/dvb-frontends/dibx000_common.c:213                                return 0;
drivers/media/dvb-frontends/dibx000_common.c:217                                return 0;
drivers/media/dvb-frontends/dibx000_common.c:235                                return 0;
drivers/media/dvb-frontends/dibx000_common.c:239                                return 0;
drivers/hv/channel.c:898                return 0;
drivers/hv/channel.c:944                return 0;

A bunch of those look suspicious but I don't know the subsystems well
enough to be sure.  Can you check the last two?

dan carpenter

Thanks, Dan!

After I checked the code, I think there is no issue for the last two:
in the case of "if (ret != 0) return 0;",  the output parameter buffer_actual_len is
zero and it is explicitly checked by the callers.
This may seem not natural and I think we can improve it in future.

BTW, at the end of vmbus_recvpacket(), the "return 0;" should be "return ret;", but
since the output parameter buffer_actual_len is checked by the callers, I think it's
OK for now.

Thanks,
-- Dexuan