From mboxrd@z Thu Jan 1 00:00:00 1970 From: AceLan Kao Subject: Re: [PATCH] acer-wmi: support Acer Aspire/eMachines 4739z Date: Thu, 8 Sep 2011 14:38:18 +0800 Message-ID: References: <1315380728-22031-1-git-send-email-acelan.kao@canonical.com> <1315387833.8674.62.camel@linux-s257.site> <1315455875.8674.110.camel@linux-s257.site> Mime-Version: 1.0 Content-Type: text/plain; charset=Big5 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from mail-qw0-f42.google.com ([209.85.216.42]:36331 "EHLO mail-qw0-f42.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752489Ab1IHGiV convert rfc822-to-8bit (ORCPT ); Thu, 8 Sep 2011 02:38:21 -0400 Received: by qwi4 with SMTP id 4so406206qwi.1 for ; Wed, 07 Sep 2011 23:38:20 -0700 (PDT) In-Reply-To: <1315455875.8674.110.camel@linux-s257.site> Sender: platform-driver-x86-owner@vger.kernel.org List-ID: To: joeyli Cc: platform-driver-x86@vger.kernel.org, mjg@redhat.com Hi, 2011/9/8 joeyli : > =A9=F3 =A5|=A1A2011-09-08 =A9=F3 10:30 +0800=A1AAceLan Kao =B4=A3=A8=EC= =A1G >> Hi, >> >> 2011/9/7 joeyli : >> > =A9=F3 =A4T=A1A2011-09-07 =A9=F3 15:32 +0800=A1AAceLan Kao =B4=A3=A8= =EC=A1G >> >> There are 2 kind of 4739Z, the one that vendor name is ACER sents= out >> >> the correct brightness key events, so we use the >> > >> > Current brightness key events means the Fn key direct send out key= code >> > but not from wmi notify event? >> Currently, most of the machines send out the correctly brightness ke= y >> code from EC, >> so we don't have to handle them. The video ACPI will capture the >> events and adjust >> the brightness. >> Acer Aspire 4739Z sends out the correct brightness key code as other >> machines do. >> Acer eMachines 4739Z doesn't sends out correct key codes, it sends o= ut >> meaningless >> key codes, so we have to capture the brightness event from WMI and r= emapping to > > What's the number of meaningless key codes? Maybe we should remap it = in > udev but not in acer-wmi driver. brightness up is - brightness down is - The reason why we shouldn't remap the keys in udev is that when it comes to follow the Acer WMI spec., the BIOS engineer won't care about which key code will be submitted while the brightness key be pressed. So, the key codes might change if there is a BIOS upgrade. > >> the correct key code and raise the correct key events to ACPI. >> > > Transfer wmi notify event to key event is fine, but keep and maintain > the current brightness level in acer-wmi is not a good idea, because > BIOS also maintain a BRTL flag in DSDT. > If there have other userland application touch brightness level throu= gh > _BCM, then will have risk causes the value out of sync. Yes, that's a problem. I don't have any idea to overcome it right now. > >> > >> >> quirk_acer_travelmate_2490. And another one with eMachines as it'= s >> >> vendor name follows Acer WMI spec. and doesn't sent out brightnes= s key >> >> events, so I added a new quirk, quirk_acer_aspire_4739z, for it, = and >> >> handle the key events in acer_wmi_notify(). >> >> >> >> BTW, we need other key event types that defined in Acer WMI spec.= , so >> >> I added them in acer_wmi_event_ids enumerate. >> >> >> >> Signed-off-by: AceLan Kao >> >> --- >> >> drivers/platform/x86/acer-wmi.c | 69 +++++++++++++++++++++++++= +++++++++++-- >> >> 1 files changed, 65 insertions(+), 4 deletions(-) >> >> >> >> diff --git a/drivers/platform/x86/acer-wmi.c b/drivers/platform/x= 86/acer-wmi.c >> >> index df730c9..257c7f2 100644 >> >> --- a/drivers/platform/x86/acer-wmi.c >> >> +++ b/drivers/platform/x86/acer-wmi.c >> >> @@ -94,6 +94,10 @@ MODULE_ALIAS("wmi:676AA15E-6A47-4D9F-A2CC-1E6D= 18D14026"); >> >> >> >> enum acer_wmi_event_ids { >> >> WMID_HOTKEY_EVENT =3D 0x1, >> >> + WMID_HOTKEY_BREAK_EVENT =3D 0x02, >> >> + WMID_GRAPHIC_SWITCH_EVENT =3D 0x03, >> > >> > Please change to 0x2 and 0x3 for consistency in this enum. >> Okay, will do. >> > > I read your new patch, thank's for your change. > >> > >> >> + WMID_BRIGHTNESS_CHANGE_EVENT =3D 0x4, >> >> + WMID_SENSOR_EVENT =3D 0x5, >> >> }; >> >> >> >> static const struct key_entry acer_wmi_keymap[] =3D { >> >> @@ -112,9 +116,9 @@ static const struct key_entry acer_wmi_keymap= [] =3D { >> >> {KE_IGNORE, 0x45, {KEY_STOP} }, >> >> {KE_IGNORE, 0x48, {KEY_VOLUMEUP} }, >> >> {KE_IGNORE, 0x49, {KEY_VOLUMEDOWN} }, >> >> - {KE_IGNORE, 0x61, {KEY_SWITCHVIDEOMODE} }, >> >> - {KE_IGNORE, 0x62, {KEY_BRIGHTNESSUP} }, >> >> - {KE_IGNORE, 0x63, {KEY_BRIGHTNESSDOWN} }, >> >> + {KE_KEY, 0x61, {KEY_SWITCHVIDEOMODE} }, >> >> + {KE_KEY, 0x62, {KEY_BRIGHTNESSUP} }, >> >> + {KE_KEY, 0x63, {KEY_BRIGHTNESSDOWN} }, >> > >> > Why enable 0x61, 0x62 and 0x63? Does it for Aspire 4739z or eMachi= nes >> > 4739z? >> It's only for eMachines now and only be activated when quirk->bright= ness =3D=3D 2. >> >> > Why choice 0x61, 0x62 and 0x63? >> > 0x61, 0x62 and 0x63 are from your machine's wmi notify event resul= t? >> They used to be those values, so I just enable them. >> And indeed, Acer Aspire 4739Z reports those key codes, 0x62, 0x63, w= hile >> brightness key be pressed, so I think other machines might report th= e same key >> codes, but they don't need to handle that. >> acer-wmi doesn't handle brightness change event, so there will be an >> error message >> comes from acer_wmi_notify(), says "Unknown function number - 3 - xx= ", that's >> why those value became ignored. >> >> We can choose any value we like other than, 0x62, 0x63, since we don= 't care what >> key codes are reported, I just need a mapping entry here, so that > > But, why you enable 0x61? It not related to brightness control, which > machine use it? It's for video toggle, both Aspire and eMachines 4739Z need that key ev= ent. > > On the other hand, > I will implement a function to fill the keymap when acer-wmi probe, t= hat > avoid some problem from we hard-code keymap. > >> I can use sparse_keymap_report_entry() later in acer_wmi_notify() >> function to report >> the correct brightness key events. >> >> > >> > But I didn't see the result from your following code.=3D >> In acer_wmi_notify() function, I use the values, 0x62, 0x63, directl= y >> to remap to the >> correct brightness key code. >> >> > >> >> {KE_KEY, 0x64, {KEY_SWITCHVIDEOMODE} }, /* Display Switch *= / >> >> {KE_IGNORE, 0x81, {KEY_SLEEP} }, >> >> {KE_KEY, 0x82, {KEY_TOUCHPAD_TOGGLE} }, /* Touch Pad On/Off= */ >> >> @@ -131,6 +135,15 @@ struct event_return_value { >> >> u32 reserved; >> >> } __attribute__((packed)); >> >> >> >> + >> >> +struct brightness_event_return_value { >> >> + u8 function; >> >> + u8 brightness_level1; >> >> + u8 brightness_level2; >> > >> > Please don't use brightness_level"1" or level"2", it confused to m= e or >> > other code reviewer and nobody can remember the definition after a >> > couple of months. >> > Please try to give them more meaningful names. >> Got it. >> > > thank's > >> > >> >> + u8 reserved; >> >> + u32 reserved2; >> >> +} __attribute__((packed)); >> >> + >> >> /* >> >> * GUID3 Get Device Status device flags >> >> */ >> >> @@ -199,6 +212,7 @@ enum interface_flags { >> >> #define ACER_DEFAULT_THREEG 0 >> >> >> >> static int max_brightness =3D 0xF; >> >> +static int cur_brightness=3D 9; >> >> >> > >> > Why should you control the current brightness level but not contro= l by >> > _BCL/_BCM in DSDT? >> The value here is just recording the current brightness. >> Because we don't know which key the user pressed, we just know the >> brightness change event happened and what the new brightness value i= s, >> so we should record the current brightness value and then compare to= the >> new value, so that we could know the brightness up or down key be pr= essed. >> >> Although, in the WMI spec., the brightness value is defined from 0x0= ~ 0x64, >> what I observed is 0 ~ 9, so I assume that the default brightness is= 9, the max. >> The value could be wrong for the first time adjusting the brightness= , that's not >> perfect, but acceptable I think. >> > > Like I said, keep current brightness acer-wmi is not good idea and ha= ve > risk to conflict with acpi video driver if there have other userland > application change brightness level through _BCM. You are right, it's better to reconsider the implementation. > >> > >> >> static int mailled =3D -1; >> >> static int brightness =3D -1; >> >> @@ -310,6 +324,10 @@ static struct quirk_entry quirk_lenovo_ideap= ad_s205 =3D { >> >> .wireless =3D 3, >> >> }; >> >> >> >> +static struct quirk_entry quirk_acer_aspire_4739z=3D { >> >> + .brightness =3D 2, >> >> +}; >> >> + >> >> /* The Aspire One has a dummy ACPI-WMI interface - disable it */ >> >> static struct dmi_system_id __devinitdata acer_blacklist[] =3D { >> >> { >> >> @@ -368,6 +386,24 @@ static struct dmi_system_id acer_quirks[] =3D= { >> >> }, >> >> { >> >> .callback =3D dmi_matched, >> >> + .ident =3D "Acer Aspire 4739Z", >> >> + .matches =3D { >> >> + DMI_MATCH(DMI_SYS_VENDOR, "Acer"), >> >> + DMI_MATCH(DMI_PRODUCT_NAME, "AS4739Z"), >> >> + }, >> >> + .driver_data =3D &quirk_acer_travelmate_2490, >> >> + }, >> > >> > Why "Acer Aspire 4739Z" set to quirk the same with travelmate 2490= ? >> > There only have a mailled capability flag in this quirk, why you u= sed >> > it? >> > Does that because Aspire 4739Z also have mail hotkey button? >> > >> > If the answer is NO, then don't need add a new quirk. >> Sorry, I missunderstanding it, there is no mailled on it, I'll remov= e it. >> > > I read your new patch, please direct remove whole Acer Aspire 4739Z > quirk block. Do you have any good reason to add the quirk block? Acer Aspire 4739Z needs the video toggling function and the key event is {KE_KEY, 0x61, {KEY_SWITCHVIDEOMODE} }, that's why we need it's quirk here. > >> >> + { >> >> + .callback =3D dmi_matched, >> >> + .ident =3D "Acer eMachines 4739Z", >> >> + .matches =3D { >> >> + DMI_MATCH(DMI_SYS_VENDOR, "eMachines"), >> >> + DMI_MATCH(DMI_PRODUCT_NAME, "AS4739Z"), >> >> + }, >> >> + .driver_data =3D &quirk_acer_aspire_4739z, >> >> + }, >> >> + { >> >> + .callback =3D dmi_matched, >> >> .ident =3D "Acer Aspire 5100", >> >> .matches =3D { >> >> DMI_MATCH(DMI_SYS_VENDOR, "Acer"), >> >> @@ -1446,7 +1482,7 @@ static void acer_rfkill_exit(void) >> >> static ssize_t show_bool_threeg(struct device *dev, >> >> struct device_attribute *attr, char *buf) >> >> { >> >> - u32 result; \ >> >> + u32 result; >> >> acpi_status status; >> >> >> >> pr_info("This threeg sysfs will be removed in 2012" >> >> @@ -1501,6 +1537,7 @@ static void acer_wmi_notify(u32 value, void= *context) >> >> struct acpi_buffer response =3D { ACPI_ALLOCATE_BUFFER, NUL= L }; >> >> union acpi_object *obj; >> >> struct event_return_value return_value; >> >> + struct brightness_event_return_value brightness_return_valu= e; >> >> acpi_status status; >> >> u16 device_state; >> >> const struct key_entry *key; >> >> @@ -1531,6 +1568,7 @@ static void acer_wmi_notify(u32 value, void= *context) >> >> >> >> switch (return_value.function) { >> >> case WMID_HOTKEY_EVENT: >> >> + case WMID_HOTKEY_BREAK_EVENT: > > Why add BREAK EVENT here? for which key? > Does it possible causes other machine run the same function key twice= ? =46or video toggling, {KE_KEY, 0x61, {KEY_SWITCHVIDEOMODE} }, And yes, you are right, that might lead to the problem that switch the output twice. I'll add one more check for this key. > >> >> device_state =3D return_value.device_state; >> >> pr_debug("device state: 0x%x\n", device_state); >> >> >> >> @@ -1558,6 +1596,29 @@ static void acer_wmi_notify(u32 value, voi= d *context) >> >> 1, true); >> >> } >> >> break; >> >> + case WMID_BRIGHTNESS_CHANGE_EVENT: >> >> + // to avoid the brightness event be handled twice >> >> + if( quirks->brightness !=3D 2) >> >> + break; >> >> + brightness_return_value =3D *((struct brightness_ev= ent_return_value *)&return_value); >> >> + /* brightness is decreasing */ >> >> + if( cur_brightness > brightness_return_value.bright= ness_level1) >> >> + key =3D sparse_keymap_entry_from_scancode(a= cer_wmi_input_dev, 0x63); >> >> + /* brightness is increasing */ >> >> + else if( cur_brightness < brightness_return_value.b= rightness_level1) >> >> + key =3D sparse_keymap_entry_from_scancode(a= cer_wmi_input_dev, 0x62); >> >> + /* brightness is already at the bottom */ >> >> + else if( cur_brightness =3D=3D 0) >> >> + key =3D sparse_keymap_entry_from_scancode(a= cer_wmi_input_dev, 0x63); >> >> + /* brightness is already at the top */ >> >> + else if( cur_brightness =3D=3D 9) >> >> + key =3D sparse_keymap_entry_from_scancode(a= cer_wmi_input_dev, 0x62); >> >> + >> >> + sparse_keymap_report_entry(acer_wmi_input_dev, key,= 1, true); >> >> + cur_brightness=3D return_value.key_num; >> >> + break; >> > >> > The same, >> > why should you control the brightness level number in acer-wmi dri= ver? >> > Does it not control by acpi video driver and _BCL/_BCM in DSDT? >> > >> > And, why choice 0x62, 0x63? The above 2 number is not from your wm= i >> > notify result, why use them? >> The problem is that the BIOS/EC doesn't send out the correct brightn= ess >> key events, the BIOS engineer just choose non-used values for those = keys >> that are defined in Acer WMI spec. Because they think those key even= ts >> should be captured from WMI, not from BIOS/EC. >> So that, we have to read the brightness change event from WMI key ev= ents. >> And then use the ACPI way to adjust the brightness. >> > > After reviewed your DSDT, I thought _Q8E and _Q8F are mapping to your > brightness up/down Fn key, please enable acpi debug message to confir= m > it. > > And, in _Q8E/Q8F: > > Method (_Q8E, 0, NotSerialized) /* Brightness up ? */ > { > If (CondRefOf (HBRT)) > { > HBRT (0x03) > } > > If (IGDS) > { > If (And (0x04, DSEN)) /* when DSEn is 0x04 i= s 100 */ > { > BRTN (0x86) /* ACPI_VIDEO_NOTIFY_I= NC_BRIGHTNESS 0x86 */ > } > > The machine emit standard ACPI_VIDEO_NOTIFY_INC_BRIGHTNESS event to > video device when IDGS is true and DSEN is 0x04. > > Please kindly check with BIOS team for what's the definition with IGD= S > and DSEN, and why the DSEN should set to 100 (looks likes it need > disable automatically control the brightness level in BIOS). > > This is a good chance to ask BIOS team to know more useful informatio= n > to us. I'm sorry that the BIOS team won't change anything for us. Acer Aspire 4739Z is already an shipping product, and Acer eMachines 4739Z only change the BIOS bootup logo and vendor name (and maybe mixed up some key codes), so they won't change anything from BIOS to avoid inducing problems to the product. And the ACPI log I gave you a month ago is Acer Aspire 4739Z, so the key codes are correct. The eMachine I have now is broken, I'm trying to reinstall it and dump the eMachines' ACPI log to you later. > > I don't want maintain brightness level in acer-wmi, we can transfer t= he > wmi event to key event, but maintain current brightness level is not > good and dirty. Agree, I need more time to work out other solution. Talk to you later. > >> > >> >> + case WMID_GRAPHIC_SWITCH_EVENT: >> >> + case WMID_SENSOR_EVENT: >> >> default: >> >> pr_warn("Unknown function number - %d - %d\n", >> >> return_value.function, return_value.key_num= ); >> > >> > > > > Thank's a lot! > Joey Lee > > -- > To unsubscribe from this list: send the line "unsubscribe platform-dr= iver-x86" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > --=20 Chia-Lin Kao(AceLan) http://blog.acelan.idv.tw/ E-Mail: acelan.kaoATcanonical.com (s/AT/@/)