Hi James, > The following patch adds 2.6 kernel support to the lm92 driver from > sensors-2.9.0. Tested with linux-2.6.11-rc5 on ppc32. Damn. I have done the exact same work one week ago... I have set a link to my patch on our "Supported Devices" page, but maybe I should have advertised more about it. I'm really sorry we duplicated the work. > - fix endian bugs - use cpu_to_le16(), not swab16() I do not think there actually is a bug. The SMBus specs say that word values are transmitted LSB first (hint: SMBus was developed by Intel). The LM92 (and every other hardware monitoring chip I know of, for that matter) transmits MSB first. Rationale for this is that it allows people not interested in decimals to only read the MSB, or so I read. So the swab16() is there for this reason, which isn't related to the CPU of the host system in any way. If you really needed to do that change for the lm92 driver to work properly, then I would suspect a bug in the i2c bus driver you are using. Which is it, BTW? > - if max6635 is detected, show a different device name in sysfs I had done this in the first place too, but now I am not sure it is a good idea. Doing this requires adding some code to libsensors, sensors and about every program relying on libsensors. I don't think it's worth the effort, especially since the LM92 doesn't seem to be found on any PC motherboard. I suspect it is actually only found on homebrew designs. Is it your case? The MAX6635 is really a clone of the LM92, so I see little benefit in having a separate prefix. My patch for the new lm92 driver is here: http://jdelvare.net1.nerim.net/sensors/linux-2.6.11-rc4-bk5-i2c-lm92.diff It is known to be broken, in particular setting limits doesn't work. I have fixed almost everything yesterday. I'll finish the job this evening, then I'll publish the new version, and would be very happy if you could test it. Of course another possibility would be that we go with your patch. However, since I have a great experience with hardware monitoring drivers in Linux 2.6, it's likely that there will be less fixes required if we start from my version. Also, for the lm92 I did not really port the original driver but merely rewrote it, because the original driver was very different from the other ones, making code maintainance over time much harder. Thanks, -- Jean Delvare
