lkml.org 
[lkml]   [2016]   [Apr]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [EXT] Re: [PATCH RESEND 3/5] USB: serial: cp210x: Added comments to CRTSCT flag code.
On Wed, Apr 27, 2016 at 08:06:32PM +0000, Konstantin Shkolnyy wrote:
> > -----Original Message-----
> > From: Johan Hovold [mailto:jhovold@gmail.com] On Behalf Of Johan Hovold
> > Sent: Tuesday, April 26, 2016 02:26
> > To: Konstantin Shkolnyy
> > Cc: Johan Hovold; Konstantin Shkolnyy; linux-usb@vger.kernel.org; linux-
> > kernel@vger.kernel.org
> > Subject: Re: [EXT] Re: [PATCH RESEND 3/5] USB: serial: cp210x: Added
> > comments to CRTSCT flag code.
> >
> > On Mon, Apr 25, 2016 at 06:09:01PM +0000, Konstantin Shkolnyy wrote:
> > > I was planning to define all these bits in a separate future patch.
> > > Would you rather prefer the magic numbers defined before fixing the
> > bugs?
> >
> > Fixing the RTS bug (patch 1), which is the only "real" bug, should be
> > done before adding defines, and fixing and cleaning up the rest.
> >
> > > I guess I can do that. Is something like this acceptable?
> > >
> > > /* CP210X_GET_FLOW/CP210X_SET_FLOW read/write these 0x10 bytes */
> > > struct cp210x_flow_ctl {
> > > u8 SERIAL_DTR_MASK : 2; /* byte 0 */
> > > u8 : 1;
> > > u8 SERIAL_CTS_HANDSHAKE : 1;
> > > u8 SERIAL_DSR_HANDSHAKE : 1;
> > > u8 SERIAL_DCD_HANDSHAKE : 1;
> > > u8 SERIAL_DSR_SENSITIVITY : 1;
> > > u8 : 1;
> > > u8; /* byte 1 */
> > > u8; /* byte 2 */
> > > u8; /* byte 3 */
> > > u8 SERIAL_AUTO_TRANSMIT : 1; /* byte 4 */
> > > u8 SERIAL_AUTO_RECEIVE : 1;
> > > u8 SERIAL_ERROR_CHAR : 1;
> > > u8 SERIAL_NULL_STRIPPING : 1;
> > > u8 SERIAL_BREAK_CHAR : 1;
> > > u8 : 1;
> > > u8 SERIAL_RTS_MASK : 2;
> > > u8; /* byte 5 */
> > > u8; /* byte 6 */
> > > u8 : 7; /* byte 7 */
> > > u8 SERIAL_XOFF_CONTINUE : 1;
> > > __le32 ulXonLimit;
> > > __le32 ulXoffLimit;
> > > } __packed;
> >
> > No, shouldn't rely on the layout of bitfields. Define masks and shifts
> > as needed and the message structure as
> >
> > struct cp210x_flow_ctl {
> > __le32 ulControlHandshake;
> > __le32 ulFlowReplace;
> > __le32 ulXonLimit;
> > __le32 ulXoffLimit;
> > };
> >
> > that is, as per AN571.
>
> OK, from searching www I see that bitfields have bad reputation for
> unclear reasons, so I guess it's now easier to avoid them.
> But doing it like you suggest, instead of splitting it to bytes, would
> complicate the code with endian conversions.
> Is there a reason for this other than making it identical to the spec?\

Staying aligned with the specification is usually a good idea. That also
became apparent when reviewing these patching and trying to match up the
magic constants with the spec.

The endian conversions should not need to complicate things that much.
Get the values using le32_to_cpu, manipulate the bits in a u32, and
store them back using cpu_to_le32.

Thanks,
Johan

\
 
 \ /
  Last update: 2016-04-28 13:21    [W:0.079 / U:17.788 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site