Hi Keith, Nice to see you around again :) > I've finally gotten around to upgrading my kernel, and the immediate > resume problem also exists in kernel 2.6.3. I've tracked down the > problem to the line > > w83781d_write_value(client, W83781D_REG_IRQ, 0x41); Interesting. This demonstrates (once again) that reseting chips and then setting registers to artibrary values is *bad. > in the file w83781d.c, introduced in lm_sensors 2.6.4; with this line > commented out, suspend works fine again. I'd like to refine the fix a bit, and then commit it to Linux 2.6 and our CVS lm_sensors2 repository. Could you please tell me: 1* Which chipset you have exactly. 2* Which value is present in the register 0x4C before you load w83781d for the first time. (Load the bus driver and use i2cdump to dump the chip contents.) 3* The value present in the same register after you load and unload the w83781d driver *with your modification* (i.e. we'll get the value resulting from the reset operation). With this information, I should be able to adapt the write to the register instead of plain removing it. Thanks. -- Jean Delvare http://www.ensicaen.ismra.fr/~delvare/
