Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 Link: https://bugs.debian.org/1033058 Fixes: 241d2fb56a18 ("of: Make OF framebuffer device names unique") Cc: stable@vger.kernel.org Signed-off-by: Cyril Brulebois cyril@debamax.com --- drivers/video/fbdev/offb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/video/fbdev/offb.c b/drivers/video/fbdev/offb.c index b97d251d894b..6264c7184457 100644 --- a/drivers/video/fbdev/offb.c +++ b/drivers/video/fbdev/offb.c @@ -698,7 +698,7 @@ MODULE_DEVICE_TABLE(of, offb_of_match_display);
static struct platform_driver offb_driver_display = { .driver = { - .name = "of-display", + .name = "of-display.0", .of_match_table = offb_of_match_display, }, .probe = offb_probe_display,
Hi
looping in as well the regressions list (hoping not doing any mistake with the regzbot commands):
On Wed, Apr 12, 2023 at 11:55:08AM +0200, Cyril Brulebois wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 Link: https://bugs.debian.org/1033058 Fixes: 241d2fb56a18 ("of: Make OF framebuffer device names unique") Cc: stable@vger.kernel.org Signed-off-by: Cyril Brulebois cyril@debamax.com
drivers/video/fbdev/offb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/video/fbdev/offb.c b/drivers/video/fbdev/offb.c index b97d251d894b..6264c7184457 100644 --- a/drivers/video/fbdev/offb.c +++ b/drivers/video/fbdev/offb.c @@ -698,7 +698,7 @@ MODULE_DEVICE_TABLE(of, offb_of_match_display); static struct platform_driver offb_driver_display = { .driver = {
.name = "of-display",
.name = "of-display.0",
#regzbot ^introduced 241d2fb56a18 #regzbot title: Open Firmware framebuffer cannot find of-display #regzbot link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 #regzbot link: https://lore.kernel.org/all/20230412095509.2196162-1-cyril@debamax.com/T/#m3... #regzbot link: https://bugs.debian.org/1033058
Regards, Salvatore
On 16.04.23 14:34, Salvatore Bonaccorso wrote:
On Wed, Apr 12, 2023 at 11:55:08AM +0200, Cyril Brulebois wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
[...] #regzbot ^introduced 241d2fb56a18 #regzbot title: Open Firmware framebuffer cannot find of-display #regzbot link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 #regzbot link: https://lore.kernel.org/all/20230412095509.2196162-1-cyril@debamax.com/T/#m3... #regzbot link: https://bugs.debian.org/1033058
No reply to my status inquiry[1] a few weeks ago, so I have to assume nobody cares anymore. If somebody still cares, holler!
#regzbot inconclusive: no answer to a status inquiry #regzbot ignore-activity
[1] https://lore.kernel.org/lkml/d1aee7d3-05f6-0920-b8e1-4ed5cf3f9f70@leemhuis.i...
Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) -- Everything you wanna know about Linux kernel regression tracking: https://linux-regtracking.leemhuis.info/about/#tldr If I did something stupid, please tell me, as explained on that page.
Hi
Am 15.06.23 um 15:03 schrieb Linux regression tracking (Thorsten Leemhuis):
On 16.04.23 14:34, Salvatore Bonaccorso wrote:
On Wed, Apr 12, 2023 at 11:55:08AM +0200, Cyril Brulebois wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
[...] #regzbot ^introduced 241d2fb56a18 #regzbot title: Open Firmware framebuffer cannot find of-display #regzbot link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 #regzbot link: https://lore.kernel.org/all/20230412095509.2196162-1-cyril@debamax.com/T/#m3... #regzbot link: https://bugs.debian.org/1033058
No reply to my status inquiry[1] a few weeks ago, so I have to assume nobody cares anymore. If somebody still cares, holler!
I'd take a look if anyone can point me to an example of Geert's proposal.
Best regards Thomas
#regzbot inconclusive: no answer to a status inquiry #regzbot ignore-activity
[1] https://lore.kernel.org/lkml/d1aee7d3-05f6-0920-b8e1-4ed5cf3f9f70@leemhuis.i...
Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
Everything you wanna know about Linux kernel regression tracking: https://linux-regtracking.leemhuis.info/about/#tldr If I did something stupid, please tell me, as explained on that page.
Hello,
On Thu, Jun 15, 2023 at 03:06:28PM +0200, Thomas Zimmermann wrote:
Hi
Am 15.06.23 um 15:03 schrieb Linux regression tracking (Thorsten Leemhuis):
On 16.04.23 14:34, Salvatore Bonaccorso wrote:
On Wed, Apr 12, 2023 at 11:55:08AM +0200, Cyril Brulebois wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
At the time this was proposed it was said that "of-display", is wrong, and that "of-display.0" must be used for the first device instead, and if something breaks an alias can be provided.
So how does one provide an alias so that offb can find "of-display.0" as "of-display"?
Thanks
Michal
[...] #regzbot ^introduced 241d2fb56a18 #regzbot title: Open Firmware framebuffer cannot find of-display #regzbot link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 #regzbot link: https://lore.kernel.org/all/20230412095509.2196162-1-cyril@debamax.com/T/#m3... #regzbot link: https://bugs.debian.org/1033058
No reply to my status inquiry[1] a few weeks ago, so I have to assume nobody cares anymore. If somebody still cares, holler!
I'd take a look if anyone can point me to an example of Geert's proposal.
Best regards Thomas
#regzbot inconclusive: no answer to a status inquiry #regzbot ignore-activity
[1] https://lore.kernel.org/lkml/d1aee7d3-05f6-0920-b8e1-4ed5cf3f9f70@leemhuis.i...
Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
Everything you wanna know about Linux kernel regression tracking: https://linux-regtracking.leemhuis.info/about/#tldr If I did something stupid, please tell me, as explained on that page.
-- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Frankenstrasse 146, 90461 Nuernberg, Germany GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman HRB 36809 (AG Nuernberg)
On Thu, Jun 15, 2023 at 03:21:07PM +0200, Michal Suchánek wrote:
Hello,
On Thu, Jun 15, 2023 at 03:06:28PM +0200, Thomas Zimmermann wrote:
Hi
Am 15.06.23 um 15:03 schrieb Linux regression tracking (Thorsten Leemhuis):
On 16.04.23 14:34, Salvatore Bonaccorso wrote:
On Wed, Apr 12, 2023 at 11:55:08AM +0200, Cyril Brulebois wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
At the time this was proposed it was said that "of-display", is wrong, and that "of-display.0" must be used for the first device instead, and if something breaks an alias can be provided.
So how does one provide an alias so that offb can find "of-display.0" as "of-display"?
I'm not aware of any way. There isn't because device names and paths are not considered ABI. There are mechanisms for getting stable class device indices (e.g. i2c0, mmcblk0, fb0, fb1, etc.) though not implemented for fbN (and please don't add it).
In any case, this should be an easy fix. Though if "linux,opened" or "linux,boot-display" is not set, then you'd still get "of-display.0":
diff --git a/drivers/of/platform.c b/drivers/of/platform.c index 78ae84187449..e46482cef9c7 100644 --- a/drivers/of/platform.c +++ b/drivers/of/platform.c @@ -553,7 +553,7 @@ static int __init of_platform_default_populate_init(void) if (!of_get_property(node, "linux,opened", NULL) || !of_get_property(node, "linux,boot-display", NULL)) continue; - dev = of_platform_device_create(node, "of-display.0", NULL); + dev = of_platform_device_create(node, "of-display", NULL); of_node_put(node); if (WARN_ON(!dev)) return -ENOMEM;
Hi Rob,
Rob Herring robh@kernel.org (2023-06-15):
On Thu, Jun 15, 2023 at 03:21:07PM +0200, Michal Suchánek wrote:
At the time this was proposed it was said that "of-display", is wrong, and that "of-display.0" must be used for the first device instead, and if something breaks an alias can be provided.
So how does one provide an alias so that offb can find "of-display.0" as "of-display"?
I'm not aware of any way. There isn't because device names and paths are not considered ABI. There are mechanisms for getting stable class device indices (e.g. i2c0, mmcblk0, fb0, fb1, etc.) though not implemented for fbN (and please don't add it).
In any case, this should be an easy fix. Though if "linux,opened" or "linux,boot-display" is not set, then you'd still get "of-display.0":
diff --git a/drivers/of/platform.c b/drivers/of/platform.c index 78ae84187449..e46482cef9c7 100644 --- a/drivers/of/platform.c +++ b/drivers/of/platform.c @@ -553,7 +553,7 @@ static int __init of_platform_default_populate_init(void) if (!of_get_property(node, "linux,opened", NULL) || !of_get_property(node, "linux,boot-display", NULL)) continue;
dev = of_platform_device_create(node, "of-display.0", NULL);
dev = of_platform_device_create(node, "of-display", NULL); of_node_put(node); if (WARN_ON(!dev)) return -ENOMEM;
I've just replaced my clueless workaround with this patch on top of the kernel found in Debian 12 (Bookworm), i.e. 6.1.27 at this point, and it indeed fixes the black screen problem in the installer's context.
I didn't run a full installation to check whether this kernel is also fine after rebooting into the installed system, but as far as I understood for the original bug report[1], it wasn't affected in the first place.
1. https://bugs.debian.org/1033058
Will somebody else pick up the torch from here, and submit that for inclusion in master? Or should I re-submit the above patch on my own?
I see my Debian colleagues have already pushed an updated v6.4-rc6 in experimental, so it should be rather easy to combine checking latest master with the distribution's packaging. Once that's done, I'm quite familiar with building an updated installer image on top of it…
Thanks,
On 6/15/23 23:19, Cyril Brulebois wrote:
Hi Rob,
Rob Herring robh@kernel.org (2023-06-15):
On Thu, Jun 15, 2023 at 03:21:07PM +0200, Michal Suchánek wrote:
At the time this was proposed it was said that "of-display", is wrong, and that "of-display.0" must be used for the first device instead, and if something breaks an alias can be provided.
So how does one provide an alias so that offb can find "of-display.0" as "of-display"?
I'm not aware of any way. There isn't because device names and paths are not considered ABI. There are mechanisms for getting stable class device indices (e.g. i2c0, mmcblk0, fb0, fb1, etc.) though not implemented for fbN (and please don't add it).
In any case, this should be an easy fix. Though if "linux,opened" or "linux,boot-display" is not set, then you'd still get "of-display.0":
diff --git a/drivers/of/platform.c b/drivers/of/platform.c index 78ae84187449..e46482cef9c7 100644 --- a/drivers/of/platform.c +++ b/drivers/of/platform.c @@ -553,7 +553,7 @@ static int __init of_platform_default_populate_init(void) if (!of_get_property(node, "linux,opened", NULL) || !of_get_property(node, "linux,boot-display", NULL)) continue;
dev = of_platform_device_create(node, "of-display.0", NULL);
dev = of_platform_device_create(node, "of-display", NULL); of_node_put(node); if (WARN_ON(!dev)) return -ENOMEM;
Michal, does that patch look correct? I don't have that hardware, but that way the boot-display gets named "of-display" while all others get "of-display.x"....
I've just replaced my clueless workaround with this patch on top of the kernel found in Debian 12 (Bookworm), i.e. 6.1.27 at this point, and it indeed fixes the black screen problem in the installer's context.
... at least it fixes the issue.
I didn't run a full installation to check whether this kernel is also fine after rebooting into the installed system, but as far as I understood for the original bug report[1], it wasn't affected in the first place.
Will somebody else pick up the torch from here, and submit that for inclusion in master? Or should I re-submit the above patch on my own?
It would be good to get some more feedback, but in general if you or Rob send me a patch I can include it in the fbdev git tree for futher testing (and if nobody else disagrees).
Helge
On Tue, Jun 20, 2023 at 08:24:34AM +0200, Helge Deller wrote:
On 6/15/23 23:19, Cyril Brulebois wrote:
Hi Rob,
Rob Herring robh@kernel.org (2023-06-15):
On Thu, Jun 15, 2023 at 03:21:07PM +0200, Michal Suchánek wrote:
At the time this was proposed it was said that "of-display", is wrong, and that "of-display.0" must be used for the first device instead, and if something breaks an alias can be provided.
So how does one provide an alias so that offb can find "of-display.0" as "of-display"?
I'm not aware of any way. There isn't because device names and paths are not considered ABI. There are mechanisms for getting stable class device indices (e.g. i2c0, mmcblk0, fb0, fb1, etc.) though not implemented for fbN (and please don't add it).
In any case, this should be an easy fix. Though if "linux,opened" or "linux,boot-display" is not set, then you'd still get "of-display.0":
diff --git a/drivers/of/platform.c b/drivers/of/platform.c index 78ae84187449..e46482cef9c7 100644 --- a/drivers/of/platform.c +++ b/drivers/of/platform.c @@ -553,7 +553,7 @@ static int __init of_platform_default_populate_init(void) if (!of_get_property(node, "linux,opened", NULL) || !of_get_property(node, "linux,boot-display", NULL)) continue;
dev = of_platform_device_create(node, "of-display.0", NULL);
dev = of_platform_device_create(node, "of-display", NULL); of_node_put(node); if (WARN_ON(!dev)) return -ENOMEM;Michal, does that patch look correct? I don't have that hardware, but that way the boot-display gets named "of-display" while all others get "of-display.x"....
Yes, that's inconsistent, and IIRC it's what I originally proposed to preserve compatibility as much as possible.
The patch went through multiple iteration by me and Rob and the final version that was merged uses the consistent but incompatible naming.
The other option would be to adapt the offb driver, presumably.
I don't understand how the device name figures in attaching the driver to the device, the only other occurence of of-display is the name of the offb and ofdrm drivers.
I've just replaced my clueless workaround with this patch on top of the kernel found in Debian 12 (Bookworm), i.e. 6.1.27 at this point, and it indeed fixes the black screen problem in the installer's context.
... at least it fixes the issue.
I would expect as much. I don't have the hardware either but keeping what was there before, and only adding on top is kind of the least disruptive way of adding new stuff.
Thanks
Michal
Linux regression tracking (Thorsten Leemhuis) regressions@leemhuis.info (2023-06-15):
No reply to my status inquiry[1] a few weeks ago, so I have to assume nobody cares anymore. If somebody still cares, holler!
I still care about a proper bugfix, for upstream and for the Debian distribution, and so does Salvatore. But fixing kernel regressions isn't my day job, so I haven't got around to working on it.
Cheers,
On 4/12/23 11:55, Cyril Brulebois wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 Link: https://bugs.debian.org/1033058 Fixes: 241d2fb56a18 ("of: Make OF framebuffer device names unique") Cc: stable@vger.kernel.org Signed-off-by: Cyril Brulebois cyril@debamax.com
I've applied the series (2 patches, one of them in drm) to the fbdev git tree.
Thanks! Helge
drivers/video/fbdev/offb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/video/fbdev/offb.c b/drivers/video/fbdev/offb.c index b97d251d894b..6264c7184457 100644 --- a/drivers/video/fbdev/offb.c +++ b/drivers/video/fbdev/offb.c @@ -698,7 +698,7 @@ MODULE_DEVICE_TABLE(of, offb_of_match_display);
static struct platform_driver offb_driver_display = { .driver = {
.name = "of-display",
.name = "of-display.0",
Hi Cyril,
CC DT
On Wed, Apr 12, 2023 at 12:05 PM Cyril Brulebois cyril@debamax.com wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 Link: https://bugs.debian.org/1033058 Fixes: 241d2fb56a18 ("of: Make OF framebuffer device names unique") Cc: stable@vger.kernel.org Signed-off-by: Cyril Brulebois cyril@debamax.com
Thanks for your patch, which is now commit 27c74ea74be805cc ("fbdev/offb: Update expected device name") in fbdev/for-next
--- a/drivers/video/fbdev/offb.c +++ b/drivers/video/fbdev/offb.c @@ -698,7 +698,7 @@ MODULE_DEVICE_TABLE(of, offb_of_match_display);
static struct platform_driver offb_driver_display = { .driver = {
.name = "of-display",
.name = "of-display.0", .of_match_table = offb_of_match_display, }, .probe = offb_probe_display,
This looks like the wrong fix for me: platform drivers' names must not contain the device index, and DT-based devices are probed using the compatible value (which is "display") instead of the node name.
I think the problem is with the of_platform_default_populate_init() function, which should create proper name@unit-address device nodes, with unique unit addresses, and with the correct compatible value.
Gr{oetje,eeting}s,
Geert
On Wed, Apr 12, 2023 at 11:55:08AM +0200, Cyril Brulebois wrote:
Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"), as spotted by Frédéric Bonnard, the historical "of-display" device is gone: the updated logic creates "of-display.0" instead, then as many "of-display.N" as required.
This means that offb no longer finds the expected device, which prevents the Debian Installer from setting up its interface, at least on ppc64el.
It might be better to iterate on all possible nodes, but updating the hardcoded device from "of-display" to "of-display.0" is confirmed to fix the Debian Installer at the very least.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217328 Link: https://bugs.debian.org/1033058 Fixes: 241d2fb56a18 ("of: Make OF framebuffer device names unique") Cc: stable@vger.kernel.org Signed-off-by: Cyril Brulebois cyril@debamax.com
Reviewed-by: Michal Suchánek msuchanek@suse.de
Thanks
Michal
drivers/video/fbdev/offb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/video/fbdev/offb.c b/drivers/video/fbdev/offb.c index b97d251d894b..6264c7184457 100644 --- a/drivers/video/fbdev/offb.c +++ b/drivers/video/fbdev/offb.c @@ -698,7 +698,7 @@ MODULE_DEVICE_TABLE(of, offb_of_match_display); static struct platform_driver offb_driver_display = { .driver = {
.name = "of-display",
.name = "of-display.0",-- 2.30.2
linux-stable-mirror@lists.linaro.org
-
Cyril Brulebois -
Geert Uytterhoeven -
Helge Deller -
Linux regression tracking (Thorsten Leemhuis) -
Michal Suchánek -
Rob Herring -
Salvatore Bonaccorso -
Thomas Zimmermann