On Wed, Oct 16, 2013 at 1:22 PM, Antti Palosaari <crope@xxxxxx> wrote: > On 16.10.2013 20:19, Michael Krufky wrote: >> >> On Wed, Oct 16, 2013 at 1:09 PM, Jean Delvare <khali@xxxxxxxxxxxx> wrote: >>> >>> Hi Michael, >>> >>> On Wed, 16 Oct 2013 13:04:42 -0400, Michael Krufky wrote: >>>> >>>> YIKES!! i2c_new_probed_device() does indeed probe the hardware -- >>>> this is unacceptable, as such an action can damage the ic. >>>> >>>> Is there some additional information that I'm missing that lets this >>>> perform an attach without probe? >>> >>> >>> Oh, i2c_new_probed_device() probes the device, what a surprise! :D >>> >>> Try, I don't know, i2c_new_device() maybe if you don't want the >>> probe? ;) >>> >>> -- >>> Jean Delvare >> >> >> OK, so to confirm that I follow correctly, one can use >> i2c_new_device() to attach the sub-driver without probing, and the >> line that ensures that the correct sub-driver gets attached is >> "strlcpy(info.type, "e4000", I2C_NAME_SIZE);" ?? >> >> We're matching based on a string? I think that's kinda yucky, but if >> that's what we're doing in i2c nowadays then I'm OK with it. >> >> If not, what prevents the wrong sub-driver from attaching to a device? >> ...or conversely, how does the right sub-driver know which device to >> attach to? > > > Yes, it is that string. Driver has that string as a ID table entry. Then you > issue i2c_new_device() call with string and it attachs driver when strings > match. > > >> Again, if I'm asking "stupid questions" just point me to the >> documentation. >> >> -Mike >> > > regards > Antti OK, I get it and it does seem OK. I'm just curious what kind of impact this refactoring would have over something like the b2c2-flexcop-fe driver, who does not know which ic's to attach based on device ids, but it does probe a few frontend combinations one after another, in an order that the driver authors knew was safe. I'd imaging that we'd write some helper abstraction function to switch out the info.type string as each driver gets probed? I think that can get quite ugly, but I know that the general population thinks dvb_attach() is even uglier, so maybe this could be the right path... Wanna take a crack at b2c2-flexcop-fe? -Mike -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
