Return-path: Received: from mail-ob0-f195.google.com ([209.85.214.195]:35536 "EHLO mail-ob0-f195.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751623AbcAEQKC (ORCPT ); Tue, 5 Jan 2016 11:10:02 -0500 Subject: Re: [RFC PATCH] staging: rtl8723au: fix byte order problems To: Jes Sorensen , Julian Calaby References: <1451921374-471-1-git-send-email-sven.dziadek@gmx.de> Cc: Sven Dziadek , Greg KH , linux-wireless , "devel@driverdev.osuosl.org" , "linux-kernel@vger.kernel.org" From: Larry Finger Message-ID: <568BEAD6.2060406@lwfinger.net> (sfid-20160105_171022_912603_7FFC0DD2) Date: Tue, 5 Jan 2016 10:09:58 -0600 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=windows-1252; format=flowed Sender: linux-wireless-owner@vger.kernel.org List-ID: On 01/05/2016 09:53 AM, Jes Sorensen wrote: > Julian Calaby writes: >> Hi Sven, >> >> On Tue, Jan 5, 2016 at 2:29 AM, Sven Dziadek wrote: >>> diff --git a/drivers/staging/rtl8723au/hal/rtl8723a_cmd.c b/drivers/staging/rtl8723au/hal/rtl8723a_cmd.c >>> index 1662c03c..57f5941 100644 >>> --- a/drivers/staging/rtl8723au/hal/rtl8723a_cmd.c >>> +++ b/drivers/staging/rtl8723au/hal/rtl8723a_cmd.c >>> @@ -93,11 +93,9 @@ int FillH2CCmd(struct rtw_adapter *padapter, u8 ElementID, u32 CmdLen, >>> >>> if (h2c_cmd & BIT(7)) { >>> msgbox_ex_addr = REG_HMEBOX_EXT_0 + (h2c_box_num * EX_MESSAGE_BOX_SIZE); >>> - h2c_cmd_ex = le16_to_cpu(h2c_cmd_ex); >>> rtl8723au_write16(padapter, msgbox_ex_addr, h2c_cmd_ex); >>> } >>> msgbox_addr = REG_HMEBOX_0 + (h2c_box_num * MESSAGE_BOX_SIZE); >>> - h2c_cmd = le32_to_cpu(h2c_cmd); >>> rtl8723au_write32(padapter, msgbox_addr, h2c_cmd); >> >> While Jes has NACK'd this change, it does highlight that the h2c_cmd >> and h2c_cmd_ex variables are being used to hold both cpu-endian and >> little-endian data. A worthwhile change here might be to move the >> conversion into the function call following these lines so that they >> remain "clean". >> >> That said, I'm not sure this particular snippet of code would work on >> big-endian at all as I'm pretty sure that BIT() produces cpu-endian >> values and we know from the line you remove that h2c_cmd is >> little-endian at this point. > > I am not opposed to cleaning it up, however I hope we can just remove > all of the code down the line instead. You may want to look at the h2c > implementation I did for rtl8xxxu. I believe it does work on big-endian, > at least I know Larry has been able to test it on big-endian systems. I have tested rtl8xxxu for the RTL8192CU devices on a PowerBook G4 with a big-endian processor. I do not have any RTL8723AU devices that can be tested this way, thus I have no idea if either of the drivers work on BE. If I were to guess, I would expect the staging driver to fail and rtl8xxxu to work. Larry