Greetings,
When connected to a remote machine via the BMC KVM functionality, I am experiencing screen freezes on boot when using 6.5 stable, but not master.
The BMC on the machine in question is an ASpeed AST2600. A quick bisect shows the problematic commit to be 2fb9667 ("drm/ast: report connection status on Display Port."). This is commit f81bb0ac upstream.
I believe the problem is that the previous commit in the series e329cb5 ("drm/ast: Add BMC virtual connector") was not backported to the stable branch. As a consequence, it appears that the more accurate DP state detection is causing the kernel to believe that no display is connected, even when the BMC's virtual display is in fact in use. A cherry-pick of e329cb5 onto the stable branch resolves the issue.
Cheers, Keno
On 13/11/2023 09:34, Keno Fischer wrote:
Greetings,
When connected to a remote machine via the BMC KVM functionality, I am experiencing screen freezes on boot when using 6.5 stable, but not master.
The BMC on the machine in question is an ASpeed AST2600. A quick bisect shows the problematic commit to be 2fb9667 ("drm/ast: report connection status on Display Port."). This is commit f81bb0ac upstream.
I believe the problem is that the previous commit in the series e329cb5 ("drm/ast: Add BMC virtual connector") was not backported to the stable branch. As a consequence, it appears that the more accurate DP state detection is causing the kernel to believe that no display is connected, even when the BMC's virtual display is in fact in use. A cherry-pick of e329cb5 onto the stable branch resolves the issue.
Yes, you're right this two patches must be backported together.
I'm sorry I didn't pay enough attention, that only one of the two was picked up for the stable branch.
Is it possible to backport e329cb5 to the stable branch, or should I push it to drm-misc-fixes ?
Best regards,
(cc: gregkh)
Hi Jocelyn
Am 13.11.23 um 10:36 schrieb Jocelyn Falempe:
On 13/11/2023 09:34, Keno Fischer wrote:
Greetings,
When connected to a remote machine via the BMC KVM functionality, I am experiencing screen freezes on boot when using 6.5 stable, but not master.
The BMC on the machine in question is an ASpeed AST2600. A quick bisect shows the problematic commit to be 2fb9667 ("drm/ast: report connection status on Display Port."). This is commit f81bb0ac upstream.
I believe the problem is that the previous commit in the series e329cb5 ("drm/ast: Add BMC virtual connector") was not backported to the stable branch. As a consequence, it appears that the more accurate DP state detection is causing the kernel to believe that no display is connected, even when the BMC's virtual display is in fact in use. A cherry-pick of e329cb5 onto the stable branch resolves the issue.
Yes, you're right this two patches must be backported together.
I'm sorry I didn't pay enough attention, that only one of the two was picked up for the stable branch.
Is it possible to backport e329cb5 to the stable branch, or should I push it to drm-misc-fixes ?
I think stable, which is in cc, will pick up commit e329cb5 semi-automatically now. Otherwise, maybe ping gregkh in a few days about it.
Best regards Thomas
Best regards,
On Mon, Nov 13, 2023 at 10:49:01AM +0100, Thomas Zimmermann wrote:
(cc: gregkh)
Hi Jocelyn
Am 13.11.23 um 10:36 schrieb Jocelyn Falempe:
On 13/11/2023 09:34, Keno Fischer wrote:
Greetings,
When connected to a remote machine via the BMC KVM functionality, I am experiencing screen freezes on boot when using 6.5 stable, but not master.
The BMC on the machine in question is an ASpeed AST2600. A quick bisect shows the problematic commit to be 2fb9667 ("drm/ast: report connection status on Display Port."). This is commit f81bb0ac upstream.
I believe the problem is that the previous commit in the series e329cb5 ("drm/ast: Add BMC virtual connector") was not backported to the stable branch. As a consequence, it appears that the more accurate DP state detection is causing the kernel to believe that no display is connected, even when the BMC's virtual display is in fact in use. A cherry-pick of e329cb5 onto the stable branch resolves the issue.
Yes, you're right this two patches must be backported together.
I'm sorry I didn't pay enough attention, that only one of the two was picked up for the stable branch.
Is it possible to backport e329cb5 to the stable branch, or should I push it to drm-misc-fixes ?
I think stable, which is in cc, will pick up commit e329cb5 semi-automatically now. Otherwise, maybe ping gregkh in a few days about it.
I thikn it would be more appropriate to revert 2fb9667, as e329cb5 doesn't look like -stable material. I'll go ahead and do that.
On 13/11/2023 17:37, Sasha Levin wrote:
On Mon, Nov 13, 2023 at 10:49:01AM +0100, Thomas Zimmermann wrote:
(cc: gregkh)
Hi Jocelyn
Am 13.11.23 um 10:36 schrieb Jocelyn Falempe:
On 13/11/2023 09:34, Keno Fischer wrote:
Greetings,
When connected to a remote machine via the BMC KVM functionality, I am experiencing screen freezes on boot when using 6.5 stable, but not master.
The BMC on the machine in question is an ASpeed AST2600. A quick bisect shows the problematic commit to be 2fb9667 ("drm/ast: report connection status on Display Port."). This is commit f81bb0ac upstream.
I believe the problem is that the previous commit in the series e329cb5 ("drm/ast: Add BMC virtual connector") was not backported to the stable branch. As a consequence, it appears that the more accurate DP state detection is causing the kernel to believe that no display is connected, even when the BMC's virtual display is in fact in use. A cherry-pick of e329cb5 onto the stable branch resolves the issue.
Yes, you're right this two patches must be backported together.
I'm sorry I didn't pay enough attention, that only one of the two was picked up for the stable branch.
Is it possible to backport e329cb5 to the stable branch, or should I push it to drm-misc-fixes ?
I think stable, which is in cc, will pick up commit e329cb5 semi-automatically now. Otherwise, maybe ping gregkh in a few days about it.
I thikn it would be more appropriate to revert 2fb9667, as e329cb5 doesn't look like -stable material. I'll go ahead and do that.
Ok, that's the best thing to do, as Thomas also found that userspace can be confused by the new BMC virtual connector, so it's safer to revert.
Thanks,
linux-stable-mirror@lists.linaro.org
-
Jocelyn Falempe -
Keno Fischer -
Sasha Levin -
Thomas Zimmermann