On Tuesday, October 23, 2012 03:39:00 PM Christopher Heiny wrote:
On 10/11/2012 01:20 AM, Dmitry Torokhov wrote:
quoted
On Fri, Oct 05, 2012 at 09:09:58PM -0700, Christopher Heiny wrote:
quoted
+
+ int (*write_block)(struct rmi_phys_device *phys, u16 addr, u8 *buf,
+ int len);
+ int (*read_block)(struct rmi_phys_device *phys, u16 addr, u8 *buf,
+ int len);
+
If you declare your buffer as [const] void * instead of u8 * I think you
will be able to get rid of most of your unions.
Good point. We'll explore that. If you see it in the next patch,
you'll know it worked out.
quoted
quoted
+ * Helper fn to convert a byte array representing a 16 bit value in the
RMI + * endian-ness to a 16-bit value in the native processor's specific
endianness. + * We don't use ntohs/htons here because, well, we're not
dealing with + * a pair of 16 bit values. Casting dest to u16* wouldn't
work, because + * that would imply knowing the byte order of u16 in the
first place. The + * same applies for using shifts and masks.
+ */
+static inline u16 batohs(u8 *src)
+{
+ return src[1] * 0x100 + src[0];
+}
+
+/**
+ * Helper function to convert a 16 bit value (in host processor
endianess) to + * a byte array in the RMI endianess for u16s. See above
comment for + * why we dont us htons or something like that.
+ */
+static inline void hstoba(u8 *dest, u16 src)
+{
+ dest[0] = src % 0x100;
+ dest[1] = src / 0x100;
+}
These are not used anymore, right?
There are function drivers that we chose not to include that depend on
these. We'll drop these from rmi.h until we're ready to submit those
other function drivers.
OK, in this case, like last time we discussed, your function drivers
should use cpu_to_le16() and le16_to_cpu() instead.
Thanks.
--
Dmitry