at_keyboard: Fix keyboards that report IBM PC AT scan codes
PATCH 1/1 at_keyboard: Fix keyboards that report IBM PC AT scan codes
GRUB has support for both the IBM PC XT (Set 1) and IBM PC AT (Set 2) key scan code mappings. But the driver always hardcodes to Set 1, overwriting the mode that was queried in the grub_keyboard_controller_init() function.
This breaks keyboard mapping on machines whose firmwares are not using the Set 1. For example, this is the case in Lenovo P71 and HP Elitebook 8770W laptops that report key scan codes using Set 2.
Instead of hardcoding to Set 1, use the value that was queried before.
---
grub-core/term/at_keyboard.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
grub-core/term/at_keyboard.c +1 -1
@@ -165,7 +165,7 @@ set_scancodes (void) }#if !USE_SCANCODE_SET - ps2_state.current_set =1 ; + ps2_state.current_set = grub_keyboard_orig_set;return ;#else
Signature · 1 line
2.28.0
Hi,
Testing the proposed patch on my old Asus N53SN in Legacy failed: as soon as at_keyboard is selected, the keys are corrupted and it's impossible to do anything.
Digging into this, it appears that query_mode() returns 2 (so set2 needs to be used), but in fact internally the keycode are the ones expected by set1.
Then I tried always executing the set_scancodes() code querying the driver and it ended up working perfectly: for real set2 was then to be used.
Hence the new proposal below (Vladimir please review this, it may be "too dirty"):
commit 5e8d6963c50cd727bfb5614448f9c80743a59de1 (HEAD -> master)
Author: Renaud Métrich <[email protected]> Date: Thu Dec 3 09:43:21 2020 +0100
Force using set_scancodes() code querying the driver.
This solves issues when query_mode() returns 2 but the driver continues to send keycodes suitable for set 1 (seen on a Asus N53SN).
Resolves: rhbz#1897587
grub-core/term/at_keyboard.c +8 -1
@@ -172,7 +172,14 @@ set_scancodes (void)return ; } -#if !USE_SCANCODE_SET +/* +* Code removed: querying the driver seems to be more robust than just querying +* the mode being returned. This was tested on both: +* - a Lenovo T460s (query_mode() returning 2 and code below working) +* - a Asus N53SN (query_mode() returning 2 but code below now working: driver +* was expecting set1 keycodes) +*/ +#if 0 && !USE_SCANCODE_SET
Renaud.
Quoted 30 lines · Javier Martinez Canillas
GRUB has support for both the IBM PC XT (Set 1) and IBM PC AT (Set 2) key scan code mappings. But the driver always hardcodes to Set 1, overwriting the mode that was queried in the
grub_keyboard_controller_init()function.This breaks keyboard mapping on machines whose firmwares are not using the Set 1. For example, this is the case in Lenovo P71 and HP Elitebook 8770W laptops that report key scan codes using Set 2.
Instead of hardcoding to Set 1, use the value that was queried before.
Reported-by Renaud Métrich <[email protected]>Signed-off-by Javier Martinez Canillas <[email protected]>---
grub-core/term/at_keyboard.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)grub-core/term/at_keyboard.c +1
-1@@ -165,7 +165,7 @@ set_scancodes (void) }#if !USE_SCANCODE_SET - ps2_state.current_set =1 ; + ps2_state.current_set = grub_keyboard_orig_set;return ;#else
Quoted 6 lines · Renaud Métrich
Hi,
Testing the proposed patch on my old Asus N53SN in Legacy failed: as soon as
at_keyboardis selected, the keys are corrupted and it's impossible to do anything.Digging into this, it appears that
query_mode()returns 2 (so set2 needs to be used), but in fact internally the keycode are the ones expected by set1.
This is because the patch doesn't take into account that controller is in "translate" mode.
@Javier Martinez Canillas : Can you make your patch check whether KEYBOARD_AT_TRANSLATE is set ?
Hello Vladimir,
Quoted 4 lines · Vladimir 'phcoder' Serbinenko
Quoted 3 lines · Renaud Métrich
Hi,
Quoted 9 lines
Quoted 3 lines
Testing the proposed patch on my old Asus N53SN in Legacy failed: as soon as
at_keyboardis selected, the keys are corrupted and it's impossible to do anything.Digging into this, it appears that
query_mode()returns 2 (so set2 needs to be used), but in fact internally the keycode are the ones expected by set1.This is because the patch doesn't take into account that controller is in "translate" mode.
@Javier Martinez Canillas : Can you make your patch check whetherKEYBOARD_AT_TRANSLATEis set ?
Thanks a lot for you feedback. Renaud is working on a new patch that takes this into account and will post to the mailing list after doing some testing on the machines that where having the problem.
Best regards,
Javier
Hi Vladimir,
Thanks for the hint, this was obvious now.
Please find attached the new patch which definitely fixes the issue.
It has been tested on various hardware (see git commit details).
In a nutshell the solution is to stick to set 1 if controller is in Translate mode, and use set X otherwise, X being the queried mode.
Additionally, in controller_fini, nothing has to be restored, since nothing was changed. This fixes an issue when switching between at_keyboard, console, and at_keyboard again, in case queried set is not the actual used set.
Renaud.
Quoted 11 lines · Vladimir 'phcoder' Serbinenko
Quoted 5 lines · Renaud Métrich
Hi,
Testing the proposed patch on my old Asus N53SN in Legacy failed: as soon as
at_keyboardis selected, the keys are corrupted and it's impossible to do anything.Digging into this, it appears that
query_mode()returns 2 (so set2 needs to be used), but in fact internally the keycode are the ones expected by set1.This is because the patch doesn't take into account that controller is in "translate" mode.
@Javier Martinez Canillas : Can you make your patch check whetherKEYBOARD_AT_TRANSLATEis set ?
Quoted 4 lines
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
From eac16e0623fa99ae5a6642852321cb8f9c74b33d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Renaud=20M=C3=A9trich?= <[email protected]>
Date: Thu, 3 Dec 2020 09:13:24 +0100
Subject: [PATCH] at_keyboard: use set 1 when keyboard is in Translate mode
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
When keyboard controller acts in Translate mode (0x40 mask), then use set 1 since translation is done.
Otherwise use the mode queried from the controller (usually set 2).
Added "atkeyb" debugging messages in at_keyboard module as well.
Resolves: rhbz#1897587
Tested on:
- Asus N53SN (set 1 used)
- Dell Precision (set 1 used)
- HP Elitebook (set 2 used)
- HP G5430 (set 1 used, keyboard in XT mode!)
- Lenovo P71 & Lenovo T460s (set 2 used)
- QEMU/KVM (set 1 used)
--- grub-core/term/at_keyboard.c | 29 ++++++++++++++++++++++++----- include/grub/at_keyboard.h | 4 ++++ 2 files changed, 28 insertions(+), 5 deletions(-)
grub-core/term/at_keyboard.c +24 -5
@@ -135,20 +135,28 @@ query_mode (void)int e; e =write_mode (0 ); -if (!e) +if (!e) { +grub_dprintf ("atkeyb" ,"query_mode: write_mode(0) failed\n" );return 0 ; + }do {keyboard_controller_wait_until_ready (); ret =grub_inb (KEYBOARD_REG_DATA ); }while (ret ==GRUB_AT_ACK );/* QEMU translates the set even in no-translate mode. */ -if (ret ==0x43 || ret ==1 ) +if (ret ==0x43 || ret ==1 ) { +grub_dprintf ("atkeyb" ,"query_mode: returning 1 (ret=0x%x)\n" , ret);return 1 ; -if (ret ==0x41 || ret ==2 ) + } +if (ret ==0x41 || ret ==2 ) { +grub_dprintf ("atkeyb" ,"query_mode: returning 2 (ret=0x%x)\n" , ret);return 2 ; -if (ret ==0x3f || ret ==3 ) + } +if (ret ==0x3f || ret ==3 ) { +grub_dprintf ("atkeyb" ,"query_mode: returning 3 (ret=0x%x)\n" , ret);return 3 ; + }return 0 ; }
@@ -165,7 +173,13 @@ set_scancodes (void) }#if !USE_SCANCODE_SET - ps2_state.current_set =1 ; +if ((grub_keyboard_controller_orig &KEYBOARD_AT_TRANSLATE ) ==KEYBOARD_AT_TRANSLATE ) { +grub_dprintf ("atkeyb" ,"queried set is %d but keyboard in Translate mode, so actually in set 1\n" , grub_keyboard_orig_set); + ps2_state.current_set =1 ; + }else { +grub_dprintf ("atkeyb" ,"using queried set %d\n" , grub_keyboard_orig_set); + ps2_state.current_set = grub_keyboard_orig_set; + }return ;#else
@@ -266,6 +280,7 @@ grub_keyboard_controller_init (void) grub_keyboard_orig_set =2 ;#else grub_keyboard_controller_orig =grub_keyboard_controller_read (); +grub_dprintf ("atkeyb" ,"grub_keyboard_controller_orig = 0x%x\n" , grub_keyboard_controller_orig); grub_keyboard_orig_set =query_mode ();#endif set_scancodes ();
@@ -275,11 +290,15 @@ grub_keyboard_controller_init (void)static grub_err_t grub_keyboard_controller_fini (struct grub_term_input *term __attribute__ ((unused))) { +/* In !USE_SCANCODE_SET mode, we didn't change anything, so nothing to restore */ +#if USE_SCANCODE_SET if (ps2_state.current_set ==0 )return GRUB_ERR_NONE ; +grub_dprintf ("atkeyb" ,"restoring set %d, controller 0x%x\n" , grub_keyboard_orig_set, grub_keyboard_controller_orig);if (grub_keyboard_orig_set)write_mode (grub_keyboard_orig_set);grub_keyboard_controller_write (grub_keyboard_controller_orig); +#endif return GRUB_ERR_NONE ; }
include/grub/at_keyboard.h +4 -0
@@ -19,6 +19,10 @@#ifndef GRUB_AT_KEYBOARD_HEADER #define GRUB_AT_KEYBOARD_HEADER 1 +/* +* Refer to https://wiki.osdev.org/%228042%22_PS/2_Controller for details. +*/ +/* Used for sending commands to the controller. */ #define KEYBOARD_COMMAND_ISREADY (x) !((x) & 0x02)#define KEYBOARD_COMMAND_READ 0x20
Signature · 1 line
2.29.2
Hello,
It appears that the proposal works fine on all systems I could test except one: a HP DL380p Gen 8.
On that system, querying the set fails: the ACK bytes in write_mode(0) work perfectly, but then 0xFE ("NACK") is read, as if the keyboard didn't want to send back the set in use.
Hence, query_mode() returns 0, causing set1 to be used, but unfortunately the system expects set2 to be used.
I tried using the "resend" command, but nothing helps.
I'm hence proposing a fallback solution for this kind of hardware where the admin can add a "at_keyboard_fallback_set" environment variable in grub.cfg that would end using a specific set if query_mode() fails.
See proposed patch in attachment.
Renaud.
Quoted 8 lines · Renaud Métrich
Hi Vladimir,
Thanks for the hint, this was obvious now.
Please find attached the new patch which definitely fixes the issue.
It has been tested on various hardware (see git commit details).
Quoted 22 lines
In a nutshell the solution is to stick to set 1 if controller is in Translate mode, and use set X otherwise, X being the queried mode.
Additionally, in
controller_fini, nothing has to be restored, since nothing was changed. This fixes an issue when switching betweenat_keyboard, console, andat_keyboardagain, in case queried set is not the actual used set.Renaud.
Quoted 15 lines · Vladimir 'phcoder' Serbinenko
Quoted 5 lines · Renaud Métrich
Hi,
Testing the proposed patch on my old Asus N53SN in Legacy failed: as soon as
at_keyboardis selected, the keys are corrupted and it's impossible to do anything.Digging into this, it appears that
query_mode()returns 2 (so set2 needs to be used), but in fact internally the keycode are the ones expected by set1.This is because the patch doesn't take into account that controller is in "translate" mode.
@Javier Martinez Canillas : Can you make your patch check whetherKEYBOARD_AT_TRANSLATEis set ?_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
From baa977fa82ca10e60ad9e344e9383f42d08b13cb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Renaud=20M=C3=A9trich?= <[email protected]>
Date: Fri, 18 Dec 2020 15:39:26 +0100
Subject: [PATCH] at_keyboard: New 'at_keyboard_fallback_set' environment
variable to force the set manually upon query_mode() failing
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This seems required with HP DL380p Gen 8 systems. Indeed, with this system, we can see the following sequence:
1. controller is queried to get current configuration (returns 0x30 which is quite standard)
2. controller is queried to get the current keyboard set in used, using code 0xf0 (first part)
3. controller answers with 0xfa which means "ACK" (== ok)
4. then we send "0" to tell "we want to know which set your are supporting"
5. controller answers with 0xfa ("ACK")
6. controller should then give us 1, 2, 3 or 0x43, 0x41, 0x3f, but here it gives us 0xfe which means "NACK"
Since there seems no way to determine the current set, and in fact the controller expects set2 to be used, we need to rely on an environment variable.
Everything has been tested on this system: using 0xFE (resend command), making sure we wait for ACK in the 2 steps "write_mode", etc.
Below is litterature I used to come up with "there is no other
solution":
- https://wiki.osdev.org/%228042%22_PS/2_Controller
- http://www-ug.eecg.toronto.edu/msl/nios_devices/datasheets/PS2%20Keyboard%20Protocol.htm
- http://www.s100computers.com/My%20System%20Pages/MSDOS%20Board/PC%20Keyboard.pdf
--- grub-core/term/at_keyboard.c | 127 ++++++++++++++++++++++++++++------- 1 file changed, 101 insertions(+), 26 deletions(-)
grub-core/term/at_keyboard.c +101 -26
@@ -31,6 +31,7 @@ GRUB_MOD_LICENSE ("GPLv3+");static grub_uint8_t grub_keyboard_controller_orig;static grub_uint8_t grub_keyboard_orig_set;struct grub_ps2_state ps2_state; +static int fallback_set;static int ping_sent;
@@ -76,6 +77,8 @@ at_command (grub_uint8_t data)break ;return 0 ; } +if (i ==GRUB_AT_TRIES ) +grub_dprintf ("atkeyb" ,"at_command() timed out! (stopped after %d tries)\n" , i);return (i !=GRUB_AT_TRIES ); }
@@ -105,6 +108,21 @@ grub_keyboard_controller_read (void)#endif +static int +resend_last_result (void ) +{ +grub_uint8_t ret; +keyboard_controller_wait_until_ready (); +grub_dprintf ("atkeyb" ,"resend_last_result: sending 0xfe\n" ); +grub_outb (0xfe ,KEYBOARD_REG_DATA ); + ret =wait_ack (); +grub_dprintf ("atkeyb" ,"resend_last_result: wait_ack() returned 0x%x\n" , ret); +keyboard_controller_wait_until_ready (); + ret =grub_inb (KEYBOARD_REG_DATA ); +grub_dprintf ("atkeyb" ,"resend_last_result: read 0x%x from controller\n" , ret); +return ret; +} +static int write_mode (int mode) {
@@ -113,11 +131,17 @@ write_mode (int mode) {grub_uint8_t ack;keyboard_controller_wait_until_ready (); +grub_dprintf ("atkeyb" ,"write_mode: sending 0xf0\n" );grub_outb (0xf0 ,KEYBOARD_REG_DATA ); + ack =wait_ack (); +grub_dprintf ("atkeyb" ,"write_mode: wait_ack() returned 0x%x\n" , ack); +if (ack !=GRUB_AT_ACK ) +continue ;keyboard_controller_wait_until_ready (); +grub_dprintf ("atkeyb" ,"write_mode: sending mode %d\n" , mode);grub_outb (mode,KEYBOARD_REG_DATA ); -keyboard_controller_wait_until_ready (); ack =wait_ack (); +grub_dprintf ("atkeyb" ,"write_mode: wait_ack() returned 0x%x\n" , ack);if (ack ==GRUB_AT_NACK )continue ;if (ack ==GRUB_AT_ACK )
@@ -125,6 +149,9 @@ write_mode (int mode)return 0 ; } +if (i ==GRUB_AT_TRIES ) +grub_dprintf ("atkeyb" ,"write_mode() timed out! (stopped after %d tries)\n" , i); +return (i !=GRUB_AT_TRIES ); }
@@ -132,31 +159,66 @@ static intquery_mode (void) {grub_uint8_t ret; +grub_uint64_t endtime; +unsigned i;int e; +char *envvar; - e =write_mode (0 ); -if (!e) { -grub_dprintf ("atkeyb" ,"query_mode: write_mode(0) failed\n" ); -return 0 ; - } +for (i =0 ; i <GRUB_AT_TRIES ; i++) { +grub_dprintf ("atkeyb" ,"query_mode: sending command to controller\n" ); + e =write_mode (0 ); +if (!e) { +grub_dprintf ("atkeyb" ,"query_mode: write_mode(0) failed\n" ); +return 0 ; + } -do { -keyboard_controller_wait_until_ready (); - ret =grub_inb (KEYBOARD_REG_DATA ); - }while (ret ==GRUB_AT_ACK ); -/* QEMU translates the set even in no-translate mode. */ -if (ret ==0x43 || ret ==1 ) { -grub_dprintf ("atkeyb" ,"query_mode: returning 1 (ret=0x%x)\n" , ret); -return 1 ; - } -if (ret ==0x41 || ret ==2 ) { -grub_dprintf ("atkeyb" ,"query_mode: returning 2 (ret=0x%x)\n" , ret); -return 2 ; + endtime =grub_get_time_ms () +20 ; +do { +keyboard_controller_wait_until_ready (); + ret =grub_inb (KEYBOARD_REG_DATA ); +grub_dprintf ("atkeyb" ,"query_mode/loop: read 0x%x from controller\n" , ret); + }while ((ret ==GRUB_AT_ACK || ret ==GRUB_AT_NACK ) &&grub_get_time_ms () < endtime); +if (ret ==0xfe ) { +grub_dprintf ("atkeyb" ,"query_mode: asking controller to resend last result\n" ); + ret =resend_last_result (); +grub_dprintf ("atkeyb" ,"query_mode: read 0x%x from controller\n" , ret); + } +/* QEMU translates the set even in no-translate mode. */ +if (ret ==0x43 || ret ==1 ) { +grub_dprintf ("atkeyb" ,"query_mode: controller returned 0x%x, returning 1\n" , ret); +return 1 ; + } +if (ret ==0x41 || ret ==2 ) { +grub_dprintf ("atkeyb" ,"query_mode: controller returned 0x%x, returning 2\n" , ret); +return 2 ; + } +if (ret ==0x3f || ret ==3 ) { +grub_dprintf ("atkeyb" ,"query_mode: controller returned 0x%x, returning 3\n" , ret); +return 3 ; + } +grub_dprintf ("atkeyb" ,"query_mode: controller returned unexpected value 0x%x, retrying\n" , ret); } -if (ret ==0x3f || ret ==3 ) { -grub_dprintf ("atkeyb" ,"query_mode: returning 3 (ret=0x%x)\n" , ret); -return 3 ; + +/* +* Falling here means we tried querying and the controller returned something +* we don't understand, try to use 'at_keyboard_fallback_set' if it exists, +* otherwise return 0. +*/ + envvar =grub_env_get ("at_keyboard_fallback_set" ); +if (envvar) { + fallback_set =grub_strtoul (envvar,0 ,10 ); +if ((grub_errno) || (fallback_set <1 ) || (fallback_set >3 )) { +grub_dprintf ("atkeyb" ,"WARNING: ignoring unexpected value '%s' for '%s' variable\n" , + envvar,"at_keyboard_fallback_set" ); + fallback_set =0 ; + }else { +grub_dprintf ("atkeyb" ,"query_mode: '%s' specified in environment, returning %d\n" , +"at_keyboard_fallback_set" , fallback_set); + } +return fallback_set; } +grub_dprintf ("atkeyb" ,"WARNING: no '%s' specified in environment, returning 0\n" , +"at_keyboard_fallback_set" );return 0 ; }
@@ -165,14 +227,25 @@ set_scancodes (void) {/* You must have visited computer museum. Keyboard without scancode set knowledge. Assume XT. */ -if (!grub_keyboard_orig_set) - { -grub_dprintf ("atkeyb" ,"No sets support assumed\n" ); - ps2_state.current_set =1 ; +if (!grub_keyboard_orig_set) { +if (fallback_set) { +grub_dprintf ("atkeyb" ,"No sets support assumed but set forced to %d\n" , fallback_set); + ps2_state.current_set = fallback_set;return ; } +grub_dprintf ("atkeyb" ,"No sets support assumed, forcing to set 1\n" ); + ps2_state.current_set =1 ; +return ; + }#if !USE_SCANCODE_SET +if (fallback_set) { +grub_dprintf ("atkeyb" ,"queried set is %d but set forced to %d\n" , + grub_keyboard_orig_set, fallback_set); + ps2_state.current_set = fallback_set; +return ; + } +if ((grub_keyboard_controller_orig &KEYBOARD_AT_TRANSLATE ) ==KEYBOARD_AT_TRANSLATE ) {grub_dprintf ("atkeyb" ,"queried set is %d but keyboard in Translate mode, so actually in set 1\n" , grub_keyboard_orig_set); ps2_state.current_set =1 ;
@@ -229,6 +302,7 @@ grub_at_keyboard_is_alive (void)if (KEYBOARD_COMMAND_ISREADY (grub_inb (KEYBOARD_REG_STATUS ))) { +grub_dprintf ("atkeyb" ,"grub_at_keyboard_is_alive: controller mode before self-test: 0x%x\n" ,grub_keyboard_controller_read ());grub_outb (0xaa ,KEYBOARD_REG_STATUS ); ping_sent =1 ; }
@@ -261,6 +335,7 @@ grub_at_keyboard_getkey (struct grub_term_input *term __attribute__ ((unused)))static void grub_keyboard_controller_init (void ) { +grub_dprintf ("atkeyb" ,"initializing the controller\n" ); ps2_state.at_keyboard_status =0 ;/* Drain input buffer. */ while (1 )
@@ -282,6 +357,7 @@ grub_keyboard_controller_init (void) grub_keyboard_controller_orig =grub_keyboard_controller_read ();grub_dprintf ("atkeyb" ,"grub_keyboard_controller_orig = 0x%x\n" , grub_keyboard_controller_orig); grub_keyboard_orig_set =query_mode (); +grub_dprintf ("atkeyb" ,"grub_keyboard_orig_set = %d\n" , grub_keyboard_orig_set);#endif set_scancodes ();keyboard_controller_led (ps2_state.led_status);
@@ -329,7 +405,6 @@ grub_at_restore_hw (void)return GRUB_ERR_NONE ; } -static struct grub_term_input grub_at_keyboard_term = { .name ="at_keyboard" ,
Signature · 1 line
2.29.2
Hello Vladimir and Daniel,
Could you please let us know your thoughts on Renaud's investigation and patches? Thanks a lot for your help.
Quoted 18 lines · Renaud Métrich
Hello,
It appears that the proposal works fine on all systems I could test except one: a HP DL380p Gen 8.
On that system, querying the set fails: the ACK bytes in
write_mode(0)work perfectly, but then0xFE("NACK") is read, as if the keyboard didn't want to send back the set in use.Hence,
query_mode()returns 0, causing set1 to be used, but unfortunately the system expects set2 to be used.I tried using the "resend" command, but nothing helps.
I'm hence proposing a fallback solution for this kind of hardware where the admin can add a "
at_keyboard_fallback_set" environment variable in grub.cfg that would end using a specific set ifquery_mode()fails.
Quoted 23 lines
See proposed patch in attachment.
Renaud.
Quoted 18 lines · Renaud Métrich
Hi Vladimir,
Thanks for the hint, this was obvious now.
Please find attached the new patch which definitely fixes the issue.
It has been tested on various hardware (see git commit details).
In a nutshell the solution is to stick to set 1 if controller is in Translate mode, and use set X otherwise, X being the queried mode.
Additionally, in
controller_fini, nothing has to be restored, since nothing was changed. This fixes an issue when switching betweenat_keyboard, console, andat_keyboardagain, in case queried set is not the actual used set.Renaud.
Best regards,
Signature · 3 lines
Javier Martinez Canillas Software Engineer - Desktop Hardware Enablement Red Hat
Hi Javier and Renaud,
Sorry for late reply but I am busy with other important stuff. I will try to take a look at the patch next week.
Daniel
Quoted 53 lines · Javier Martinez Canillas
Hello Vladimir and Daniel,
Could you please let us know your thoughts on Renaud's investigation and patches? Thanks a lot for your help.
Quoted 18 lines · Renaud Métrich
Hello,
It appears that the proposal works fine on all systems I could test except one: a HP DL380p Gen 8.
On that system, querying the set fails: the ACK bytes in
write_mode(0)work perfectly, but then0xFE("NACK") is read, as if the keyboard didn't want to send back the set in use.Hence,
query_mode()returns 0, causing set1 to be used, but unfortunately the system expects set2 to be used.I tried using the "resend" command, but nothing helps.
I'm hence proposing a fallback solution for this kind of hardware where the admin can add a "
at_keyboard_fallback_set" environment variable in grub.cfg that would end using a specific set ifquery_mode()fails.Quoted 23 lines
See proposed patch in attachment.
Renaud.
Quoted 18 lines · Renaud Métrich
Hi Vladimir,
Thanks for the hint, this was obvious now.
Please find attached the new patch which definitely fixes the issue.
It has been tested on various hardware (see git commit details).
In a nutshell the solution is to stick to set 1 if controller is in Translate mode, and use set X otherwise, X being the queried mode.
Additionally, in
controller_fini, nothing has to be restored, since nothing was changed. This fixes an issue when switching betweenat_keyboard, console, andat_keyboardagain, in case queried set is not the actual used set.Renaud.
Best regards,
--
Javier Martinez Canillas
Software Engineer - Desktop Hardware Enablement
Red Hat
Hello,
Please find an updated version of the patch proposing a fallback mechanism in case grub is not able to detect the set to be used.
I has been tested on the following systems:
- Lenovo P71
- Asus N53SN
- Dell Precision
- HP Elitebook
- Lenovo T460s
- QEMU/KVM
- HP G5430
On all these above, the fallback mechanism is not used at all, set is detected correctly.
However the mechanism is necessary for HP DL380p Gen 8 systems.
Renaud.
Quoted 10 lines · Renaud Métrich
Hello,
It appears that the proposal works fine on all systems I could test except one: a HP DL380p Gen 8.
On that system, querying the set fails: the ACK bytes in
write_mode(0)work perfectly, but then0xFE("NACK") is read, as if the keyboard didn't want to send back the set in use.Hence,
query_mode()returns 0, causing set1 to be used, but unfortunately the system expects set2 to be used.I tried using the "resend" command, but nothing helps.
Quoted 37 lines
I'm hence proposing a fallback solution for this kind of hardware where the admin can add a "
at_keyboard_fallback_set" environment variable in grub.cfg that would end using a specific set ifquery_mode()fails.See proposed patch in attachment.
Renaud.
Quoted 30 lines · Renaud Métrich
Hi Vladimir,
Thanks for the hint, this was obvious now.
Please find attached the new patch which definitely fixes the issue.
It has been tested on various hardware (see git commit details).
In a nutshell the solution is to stick to set 1 if controller is in Translate mode, and use set X otherwise, X being the queried mode.
Additionally, in
controller_fini, nothing has to be restored, since nothing was changed. This fixes an issue when switching betweenat_keyboard, console, andat_keyboardagain, in case queried set is not the actual used set.Renaud.
Quoted 15 lines · Vladimir 'phcoder' Serbinenko
Quoted 5 lines · Renaud Métrich
Hi,
Testing the proposed patch on my old Asus N53SN in Legacy failed: as soon as
at_keyboardis selected, the keys are corrupted and it's impossible to do anything.Digging into this, it appears that
query_mode()returns 2 (so set2 needs to be used), but in fact internally the keycode are the ones expected by set1.This is because the patch doesn't take into account that controller is in "translate" mode.
@Javier Martinez Canillas : Can you make your patch check whetherKEYBOARD_AT_TRANSLATEis set ?_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
From 3d290e4b6129dc1e6458815635d978302002c859 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Renaud=20M=C3=A9trich?= <[email protected]>
Date: Fri, 18 Dec 2020 15:39:26 +0100
Subject: [PATCH] at_keyboard: New 'at_keyboard_fallback_set' environment
variable to force the set manually upon query_mode() failing
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This seems required with HP DL380p Gen 8 systems. Indeed, with this system, we can see the following sequence:
1. controller is queried to get current configuration (returns 0x30 which is quite standard)
2. controller is queried to get the current keyboard set in used, using code 0xf0 (first part)
3. controller answers with 0xfa which means "ACK" (== ok)
4. then we send "0" to tell "we want to know which set your are supporting"
5. controller answers with 0xfa ("ACK")
6. controller should then give us 1, 2, 3 or 0x43, 0x41, 0x3f, but here it gives us 0xfe which means "NACK"
Since there seems no way to determine the current set, and in fact the controller expects set2 to be used, we need to rely on an environment variable.
Everything has been tested on this system: using 0xFE (resend command), making sure we wait for ACK in the 2 steps "write_mode", etc.
Below is litterature I used to come up with "there is no other
solution":
- https://wiki.osdev.org/%228042%22_PS/2_Controller
- http://www-ug.eecg.toronto.edu/msl/nios_devices/datasheets/PS2%20Keyboard%20Protocol.htm
- http://www.s100computers.com/My%20System%20Pages/MSDOS%20Board/PC%20Keyboard.pdf
--- grub-core/term/at_keyboard.c | 121 +++++++++++++++++++++++++++-------- 1 file changed, 96 insertions(+), 25 deletions(-)
grub-core/term/at_keyboard.c +96 -25
@@ -31,6 +31,7 @@ GRUB_MOD_LICENSE ("GPLv3+");static grub_uint8_t grub_keyboard_controller_orig;static grub_uint8_t grub_keyboard_orig_set;struct grub_ps2_state ps2_state; +static int fallback_set;static int ping_sent;
@@ -76,6 +77,8 @@ at_command (grub_uint8_t data)break ;return 0 ; } +if (i ==GRUB_AT_TRIES ) +grub_dprintf ("atkeyb" ,"at_command() timed out! (stopped after %d tries)\n" , i);return (i !=GRUB_AT_TRIES ); }
@@ -105,6 +108,21 @@ grub_keyboard_controller_read (void)#endif +static int +resend_last_result (void ) +{ +grub_uint8_t ret; +keyboard_controller_wait_until_ready (); +grub_dprintf ("atkeyb" ,"resend_last_result: sending 0xfe\n" ); +grub_outb (0xfe ,KEYBOARD_REG_DATA ); + ret =wait_ack (); +grub_dprintf ("atkeyb" ,"resend_last_result: wait_ack() returned 0x%x\n" , ret); +keyboard_controller_wait_until_ready (); + ret =grub_inb (KEYBOARD_REG_DATA ); +grub_dprintf ("atkeyb" ,"resend_last_result: read 0x%x from controller\n" , ret); +return ret; +} +static int write_mode (int mode) {
@@ -113,11 +131,14 @@ write_mode (int mode) {grub_uint8_t ack;keyboard_controller_wait_until_ready (); +grub_dprintf ("atkeyb" ,"write_mode: sending 0xf0\n" );grub_outb (0xf0 ,KEYBOARD_REG_DATA );keyboard_controller_wait_until_ready (); +grub_dprintf ("atkeyb" ,"write_mode: sending mode %d\n" , mode);grub_outb (mode,KEYBOARD_REG_DATA );keyboard_controller_wait_until_ready (); ack =wait_ack (); +grub_dprintf ("atkeyb" ,"write_mode: wait_ack() returned 0x%x\n" , ack);if (ack ==GRUB_AT_NACK )continue ;if (ack ==GRUB_AT_ACK )
@@ -125,6 +146,9 @@ write_mode (int mode)return 0 ; } +if (i ==GRUB_AT_TRIES ) +grub_dprintf ("atkeyb" ,"write_mode() timed out! (stopped after %d tries)\n" , i); +return (i !=GRUB_AT_TRIES ); }
@@ -132,31 +156,66 @@ static intquery_mode (void) {grub_uint8_t ret; +grub_uint64_t endtime; +unsigned i;int e; +char *envvar; - e =write_mode (0 ); -if (!e) { -grub_dprintf ("atkeyb" ,"query_mode: write_mode(0) failed\n" ); -return 0 ; - } +for (i =0 ; i <GRUB_AT_TRIES ; i++) { +grub_dprintf ("atkeyb" ,"query_mode: sending command to controller\n" ); + e =write_mode (0 ); +if (!e) { +grub_dprintf ("atkeyb" ,"query_mode: write_mode(0) failed\n" ); +return 0 ; + } -do { -keyboard_controller_wait_until_ready (); - ret =grub_inb (KEYBOARD_REG_DATA ); - }while (ret ==GRUB_AT_ACK ); -/* QEMU translates the set even in no-translate mode. */ -if (ret ==0x43 || ret ==1 ) { -grub_dprintf ("atkeyb" ,"query_mode: returning 1 (ret=0x%x)\n" , ret); -return 1 ; - } -if (ret ==0x41 || ret ==2 ) { -grub_dprintf ("atkeyb" ,"query_mode: returning 2 (ret=0x%x)\n" , ret); -return 2 ; + endtime =grub_get_time_ms () +20 ; +do { +keyboard_controller_wait_until_ready (); + ret =grub_inb (KEYBOARD_REG_DATA ); +grub_dprintf ("atkeyb" ,"query_mode/loop: read 0x%x from controller\n" , ret); + }while ((ret ==GRUB_AT_ACK || ret ==GRUB_AT_NACK ) &&grub_get_time_ms () < endtime); +if (ret ==0xfe ) { +grub_dprintf ("atkeyb" ,"query_mode: asking controller to resend last result\n" ); + ret =resend_last_result (); +grub_dprintf ("atkeyb" ,"query_mode: read 0x%x from controller\n" , ret); + } +/* QEMU translates the set even in no-translate mode. */ +if (ret ==0x43 || ret ==1 ) { +grub_dprintf ("atkeyb" ,"query_mode: controller returned 0x%x, returning 1\n" , ret); +return 1 ; + } +if (ret ==0x41 || ret ==2 ) { +grub_dprintf ("atkeyb" ,"query_mode: controller returned 0x%x, returning 2\n" , ret); +return 2 ; + } +if (ret ==0x3f || ret ==3 ) { +grub_dprintf ("atkeyb" ,"query_mode: controller returned 0x%x, returning 3\n" , ret); +return 3 ; + } +grub_dprintf ("atkeyb" ,"query_mode: controller returned unexpected value 0x%x, retrying\n" , ret); } -if (ret ==0x3f || ret ==3 ) { -grub_dprintf ("atkeyb" ,"query_mode: returning 3 (ret=0x%x)\n" , ret); -return 3 ; + +/* +* Falling here means we tried querying and the controller returned something +* we don't understand, try to use 'at_keyboard_fallback_set' if it exists, +* otherwise return 0. +*/ + envvar =grub_env_get ("at_keyboard_fallback_set" ); +if (envvar) { + fallback_set =grub_strtoul (envvar,0 ,10 ); +if ((grub_errno) || (fallback_set <1 ) || (fallback_set >3 )) { +grub_dprintf ("atkeyb" ,"WARNING: ignoring unexpected value '%s' for '%s' variable\n" , + envvar,"at_keyboard_fallback_set" ); + fallback_set =0 ; + }else { +grub_dprintf ("atkeyb" ,"query_mode: '%s' specified in environment, returning %d\n" , +"at_keyboard_fallback_set" , fallback_set); + } +return fallback_set; } +grub_dprintf ("atkeyb" ,"WARNING: no '%s' specified in environment, returning 0\n" , +"at_keyboard_fallback_set" );return 0 ; }
@@ -165,14 +224,25 @@ set_scancodes (void) {/* You must have visited computer museum. Keyboard without scancode set knowledge. Assume XT. */ -if (!grub_keyboard_orig_set) - { -grub_dprintf ("atkeyb" ,"No sets support assumed\n" ); - ps2_state.current_set =1 ; +if (!grub_keyboard_orig_set) { +if (fallback_set) { +grub_dprintf ("atkeyb" ,"No sets support assumed but set forced to %d\n" , fallback_set); + ps2_state.current_set = fallback_set;return ; } +grub_dprintf ("atkeyb" ,"No sets support assumed, forcing to set 1\n" ); + ps2_state.current_set =1 ; +return ; + }#if !USE_SCANCODE_SET +if (fallback_set) { +grub_dprintf ("atkeyb" ,"queried set is %d but set forced to %d\n" , + grub_keyboard_orig_set, fallback_set); + ps2_state.current_set = fallback_set; +return ; + } +if ((grub_keyboard_controller_orig &KEYBOARD_AT_TRANSLATE ) ==KEYBOARD_AT_TRANSLATE ) {grub_dprintf ("atkeyb" ,"queried set is %d but keyboard in Translate mode, so actually in set 1\n" , grub_keyboard_orig_set); ps2_state.current_set =1 ;
@@ -261,6 +331,7 @@ grub_at_keyboard_getkey (struct grub_term_input *term __attribute__ ((unused)))static void grub_keyboard_controller_init (void ) { +grub_dprintf ("atkeyb" ,"initializing the controller\n" ); ps2_state.at_keyboard_status =0 ;/* Drain input buffer. */ while (1 )
@@ -282,6 +353,7 @@ grub_keyboard_controller_init (void) grub_keyboard_controller_orig =grub_keyboard_controller_read ();grub_dprintf ("atkeyb" ,"grub_keyboard_controller_orig = 0x%x\n" , grub_keyboard_controller_orig); grub_keyboard_orig_set =query_mode (); +grub_dprintf ("atkeyb" ,"grub_keyboard_orig_set = %d\n" , grub_keyboard_orig_set);#endif set_scancodes ();keyboard_controller_led (ps2_state.led_status);
@@ -329,7 +401,6 @@ grub_at_restore_hw (void)return GRUB_ERR_NONE ; } -static struct grub_term_input grub_at_keyboard_term = { .name ="at_keyboard" ,
Signature · 1 line
2.29.2