| Message ID | 1431192213-2904-1-git-send-email-wsa@the-dreams.de (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Sat, May 09, 2015 at 07:23:33PM +0200, Wolfram Sang wrote: > Commit 3a3dd0186f619b ("i2c/powermac: Improve detection of devices from > device-tree") added a codec device instantiation workaround > unconditionally although it is only needed for onyx. Do it conditionally > since keywest has its own codec instantiation in the sound drivers. > Thanks must go to GrowlTiger from Gentoo bug report #469132 for > bisecting. > > Reported-by: Dan DeVoto <dand1972@yahoo.com> > Signed-off-by: Wolfram Sang <wsa@the-dreams.de> > Tested-by: Dan DeVoto <dand1972@yahoo.com> > Tested-by: Mark Elliott <txlitebeer@gmail.com> > Fixes: 3a3dd0186f619b ("i2c/powermac: Improve detection of devices from device-tree") > --- > drivers/i2c/busses/i2c-powermac.c | 9 +++++---- > 1 file changed, 5 insertions(+), 4 deletions(-) > > diff --git a/drivers/i2c/busses/i2c-powermac.c b/drivers/i2c/busses/i2c-powermac.c > index 6abcf696e3594b..cad5552c0cd4e4 100644 > --- a/drivers/i2c/busses/i2c-powermac.c > +++ b/drivers/i2c/busses/i2c-powermac.c > @@ -285,7 +285,8 @@ static void i2c_powermac_add_missing(struct i2c_adapter *adap, > > static bool i2c_powermac_get_type(struct i2c_adapter *adap, > struct device_node *node, > - u32 addr, char *type, int type_size) > + u32 addr, char *type, int type_size, > + bool found_onyx) > { > char tmp[16]; > > @@ -305,8 +306,8 @@ static bool i2c_powermac_get_type(struct i2c_adapter *adap, > return true; > } > > - /* Now look for known workarounds */ > - if (!strcmp(node->name, "deq")) { > + /* Now look for known workarounds for onyx/aoa */ > + if (found_onyx && !strcmp(node->name, "deq")) { > /* Apple uses address 0x34 for TAS3001 and 0x35 for TAS3004 */ Okay, so this patch is bogus. I understand now that onyx uses another codec than TAS, so this change will regress on other machines. However, it shows that this unconditional instantiation of the TAS breaks sound on Macintoshs which still need non-aoa sound support. I assume there will be noone in the near future to convert Keywest to AOA, so we'll need to find a hackish way around this instantiation problem.
On Sun, 2015-05-10 at 20:34 +0200, Wolfram Sang wrote: > Okay, so this patch is bogus. I understand now that onyx uses another > codec than TAS, so this change will regress on other machines. > However, > it shows that this unconditional instantiation of the TAS breaks sound > on Macintoshs which still need non-aoa sound support. I assume there > will be noone in the near future to convert Keywest to AOA, so we'll > need to find a hackish way around this instantiation problem. Converting the old macs that use TAS shouldn't be *that* hard, main problem is I don't have the hardware to test... Ben.
On Mon, May 11, 2015 at 08:14:47AM +1000, Benjamin Herrenschmidt wrote: > On Sun, 2015-05-10 at 20:34 +0200, Wolfram Sang wrote: > > Okay, so this patch is bogus. I understand now that onyx uses another > > codec than TAS, so this change will regress on other machines. > > However, > > it shows that this unconditional instantiation of the TAS breaks sound > > on Macintoshs which still need non-aoa sound support. I assume there > > will be noone in the near future to convert Keywest to AOA, so we'll > > need to find a hackish way around this instantiation problem. > > Converting the old macs that use TAS shouldn't be *that* hard, main > problem is I don't have the hardware to test... Dan and Mark (on CC) have been very helpful with testing. Maybe they are also in to test "the proper solution"?
On Mon, 2015-05-11 at 09:34 +0200, wsa@the-dreams.de wrote: > On Mon, May 11, 2015 at 08:14:47AM +1000, Benjamin Herrenschmidt wrote: > > On Sun, 2015-05-10 at 20:34 +0200, Wolfram Sang wrote: > > > Okay, so this patch is bogus. I understand now that onyx uses another > > > codec than TAS, so this change will regress on other machines. > > > However, > > > it shows that this unconditional instantiation of the TAS breaks sound > > > on Macintoshs which still need non-aoa sound support. I assume there > > > will be noone in the near future to convert Keywest to AOA, so we'll > > > need to find a hackish way around this instantiation problem. > > > > Converting the old macs that use TAS shouldn't be *that* hard, main > > problem is I don't have the hardware to test... > > Dan and Mark (on CC) have been very helpful with testing. Maybe they are > also in to test "the proper solution"? Possibly, depends how many machines we can cover. I also have extremely little time, so I'm not that keen on volunteering :-) I'll *try* to have a look some time this or next week. The main problem is the difference in the way the layout of the codec etc... is reported from the DT by the firmware. Otherwise the HW is the same between pre-AOA and post-AOA really ... I do have a bunch of DT snapshots lying around, so I can try to figure out something. Dan, Mark, what machine models specifically do you have ? (compatible property in the DT pls, ie, /proc/device-tree/compatible). Cheers, Ben.
Hi, Sorry for the delay, I found this message in my spam folder. Anyway, for my machine model, cat /proc/device-tree/compatible returns: PowerBook4,3MacRISC2MacRISCPower Macintosh It's a 14 in. iBook G3 700 MHz. Regards, Dan -------------------------------------------- On Mon, 5/11/15, Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote: Subject: Re: [PATCH] i2c: powermac: don't workaround for keywest To: wsa@the-dreams.de Cc: linux-i2c@vger.kernel.org, linuxppc-dev@lists.ozlabs.org, "Dan DeVoto" <dand1972@yahoo.com>, "Mark Elliott" <txlitebeer@gmail.com> Date: Monday, May 11, 2015, 4:26 PM On Mon, 2015-05-11 at 09:34 +0200, wsa@the-dreams.de wrote: > On Mon, May 11, 2015 at 08:14:47AM +1000, Benjamin Herrenschmidt wrote: > > On Sun, 2015-05-10 at 20:34 +0200, Wolfram Sang wrote: > > > Okay, so this patch is bogus. I understand now that onyx uses another > > > codec than TAS, so this change will regress on other machines. > > > However, > > > it shows that this unconditional instantiation of the TAS breaks sound > > > on Macintoshs which still need non-aoa sound support. I assume there > > > will be noone in the near future to convert Keywest to AOA, so we'll > > > need to find a hackish way around this instantiation problem. > > > > Converting the old macs that use TAS shouldn't be *that* hard, main > > problem is I don't have the hardware to test... > > Dan and Mark (on CC) have been very helpful with testing. Maybe they are > > also in to test "the proper solution"? > Possibly, depends how many machines we can cover. I also have extremely > little time, so I'm not that keen on volunteering :-) I'll *try* to have > a look some time this or next week. > > The main problem is the difference in the way the layout of the codec > etc... is reported from the DT by the firmware. Otherwise the HW is the > same between pre-AOA and post-AOA really ... > > I do have a bunch of DT snapshots lying around, so I can try > to figure out something. > > Dan, Mark, what machine models specifically do you have ? (compatible > property in the DT pls, ie, /proc/device-tree/compatible). > > Cheers, > Ben.
diff --git a/drivers/i2c/busses/i2c-powermac.c b/drivers/i2c/busses/i2c-powermac.c index 6abcf696e3594b..cad5552c0cd4e4 100644 --- a/drivers/i2c/busses/i2c-powermac.c +++ b/drivers/i2c/busses/i2c-powermac.c @@ -285,7 +285,8 @@ static void i2c_powermac_add_missing(struct i2c_adapter *adap, static bool i2c_powermac_get_type(struct i2c_adapter *adap, struct device_node *node, - u32 addr, char *type, int type_size) + u32 addr, char *type, int type_size, + bool found_onyx) { char tmp[16]; @@ -305,8 +306,8 @@ static bool i2c_powermac_get_type(struct i2c_adapter *adap, return true; } - /* Now look for known workarounds */ - if (!strcmp(node->name, "deq")) { + /* Now look for known workarounds for onyx/aoa */ + if (found_onyx && !strcmp(node->name, "deq")) { /* Apple uses address 0x34 for TAS3001 and 0x35 for TAS3004 */ if (addr == 0x34) { snprintf(type, type_size, "MAC,tas3001"); @@ -362,7 +363,7 @@ static void i2c_powermac_register_devices(struct i2c_adapter *adap, /* Make up a modalias */ if (!i2c_powermac_get_type(adap, node, addr, - info.type, sizeof(info.type))) { + info.type, sizeof(info.type), found_onyx)) { continue; }
Commit 3a3dd0186f619b ("i2c/powermac: Improve detection of devices from device-tree") added a codec device instantiation workaround unconditionally although it is only needed for onyx. Do it conditionally since keywest has its own codec instantiation in the sound drivers. Thanks must go to GrowlTiger from Gentoo bug report #469132 for bisecting. Reported-by: Dan DeVoto <dand1972@yahoo.com> Signed-off-by: Wolfram Sang <wsa@the-dreams.de> Tested-by: Dan DeVoto <dand1972@yahoo.com> Tested-by: Mark Elliott <txlitebeer@gmail.com> Fixes: 3a3dd0186f619b ("i2c/powermac: Improve detection of devices from device-tree") --- drivers/i2c/busses/i2c-powermac.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)