xen/arm: Remove EXPERT dependancy
PATCH 1/1 xen/arm: Remove EXPERT dependancy
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI potentially taking over. Some common devices may need ACPI table support.
Presently I think it is worth removing the dependancy on CONFIG_EXPERT. I am rather tempted to add defaulting enabled, but I'm not yet confident of going that far yet.
---
I hope a popular ARM device capable of running Xen will soon be running Xen on ACPI/UEFI, but it isn't quite there yet. As such I would like to have "default y", but I don't think that is likely to be accepted yet.
--- xen/arch/arm/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/arm/Kconfig +1 -1
@@ -32,7 +32,7 @@ menu "Architecture Features" source "arch/Kconfig" config ACPI - bool "ACPI (Advanced Configuration and Power Interface) Support" if EXPERT + bool "ACPI (Advanced Configuration and Power Interface) Support" depends on ARM_64 ---help---
Signature · 9 lines
2.20.1 -- (\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \BS ( | [email protected] PGP 87145445 | ) / \_CS\ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ 8A19\___\_|_/58D2 7E3D DDF4 7BA6 <-PGP-> 41D1 B375 37D0 8714\_|_/___/5445
Hi,
Quoted 6 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI potentially taking over. Some common devices may need ACPI table support.
Presently I think it is worth removing the dependancy on
CONFIG_EXPERT.
The idea behind EXPERT is to gate any feature that is not considered to be stable/complete enough to be used in production.
I don't consider the ACPI complete because the parsing of the IORT (used to discover SMMU and GICv3 ITS) is not there yet.
I vaguely remember some issues on system using SMMU (e.g. Thunder-X) because Dom0 will try to use the IOMMU and this would break PV drivers.
Therefore I think we at least want to consider to hide SMMUs from dom0 before removing EXPERT. Ideally, I would also like the feature to be tested in Osstest.
The good news is Xen Project already has systems (e.g. Thunder-X, Softiron) that can supported ACPI. So it should hopefully be a matter to tell them to boot with ACPI rather than DT.
Cheers,
Signature · 1 line
Julien Grall
Quoted 10 lines · Julien Grall
Quoted 6 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI potentially taking over. Some common devices may need ACPI table support.
Presently I think it is worth removing the dependancy on
CONFIG_EXPERT.The idea behind EXPERT is to gate any feature that is not considered to be stable/complete enough to be used in production.
Yes, and from that point of view I don't think we want to remove EXPERT from ACPI yet. However, the idea of hiding things behind EXPERT works very well for new esoteric features, something like memory introspection or memory overcommit. It does not work well for things that are actually required to boot on the platform.
Typically ACPI systems don't come with device tree at all (RPi4 being an exception), so users don't really have much of a choice in the matter.
From that point of view, it would be better to remove EXPERT from ACPI, maybe even build ACPI by default, *but* to add a warning at boot saying something like:
"ACPI support is experimental. Boot using Device Tree if you can."
That would better convey the risks of using ACPI, while at the same time making it a bit easier for users to boot on their ACPI-only platforms.
Quoted 5 lines
I don't consider the ACPI complete because the parsing of the IORT (used to discover SMMU and GICv3 ITS) is not there yet.
I vaguely remember some issues on system using SMMU (e.g. Thunder-X) because Dom0 will try to use the IOMMU and this would break PV drivers.
I am not sure why Dom0 using the IOMMU would break PV drivers? Is it because the pagetable is not properly updated when mapping foreign pages?
Quoted 7 lines
Therefore I think we at least want to consider to hide SMMUs from dom0 before removing EXPERT. Ideally, I would also like the feature to be tested in Osstest.
The good news is Xen Project already has systems (e.g. Thunder-X, Softiron) that can supported ACPI. So it should hopefully be a matter to tell them to boot with ACPI rather than DT.
I agree that we want to keep ACPI "expert/experimental" given its current state but maybe we can find a better way to carry that message than to set EXPERT in Kconfig.
And yes, if we wanted to make ACPI less "expert/experimental" we definitely need some testing in OSSTest and any critical bugs (e.g. PV drivers not working) addressed.
Quoted 22 lines · Stefano Stabellini
Quoted 10 lines · Julien Grall
Quoted 6 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI potentially taking over. Some common devices may need ACPI table support.
Presently I think it is worth removing the dependancy on
CONFIG_EXPERT.The idea behind EXPERT is to gate any feature that is not considered to be stable/complete enough to be used in production.
Yes, and from that point of view I don't think we want to remove EXPERT from ACPI yet. However, the idea of hiding things behind EXPERT works very well for new esoteric features, something like memory introspection or memory overcommit. It does not work well for things that are actually required to boot on the platform.
Typically ACPI systems don't come with device tree at all (RPi4 being an exception), so users don't really have much of a choice in the matter.
Quoted 1 line
From that point of view, it would be better to remove EXPERT from ACPI,
Quoted 8 lines
maybe even build ACPI by default, *but* to add a warning at boot saying something like:
"ACPI support is experimental. Boot using Device Tree if you can."
That would better convey the risks of using ACPI, while at the same time making it a bit easier for users to boot on their ACPI-only platforms.
This matches my view. I was thinking about including "default y", but I felt the chances of that getting through were lower. I concur with a warning on boot being a good approach.
Signature · 4 lines
(\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \BS ( | [email protected] PGP 87145445 | ) / \_CS\ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ 8A19\___\_|_/58D2 7E3D DDF4 7BA6 <-PGP-> 41D1 B375 37D0 8714\_|_/___/5445
PATCH 1/1 xen/arm: ACPI: Remove EXPERT dependancy, default on for ARM64
In reply to Stefano StabelliniLinux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI potentially taking over. Some common devices may require ACPI table support to boot.
For devices which can boot in either mode, continue defaulting to device-tree. Add warnings about using ACPI advising users of present situation.
---
Okay, hopefully this is okay. Warning in Kconfig, warning on boot. Perhaps "default y if ARM_64" is redundant, yet if someone tries to make it possible to boot aarch32 on a ACPI machine...
I also want a date in the message. Theory is this won't be there forever, so a date is essential.
--- xen/arch/arm/Kconfig | 7 ++++++- xen/arch/arm/acpi/boot.c | 9 +++++++++ 2 files changed, 15 insertions(+), 1 deletion(-)
xen/arch/arm/Kconfig +6 -1
@@ -32,13 +32,18 @@ menu "Architecture Features" source "arch/Kconfig" config ACPI - bool "ACPI (Advanced Configuration and Power Interface) Support" if EXPERT + bool "ACPI (Advanced Configuration and Power Interface) Support" depends on ARM_64 + default y if ARM_64 ---help--- Advanced Configuration and Power Interface (ACPI) support for Xen is an alternative to device tree on ARM64. + Note this is presently EXPERIMENTAL. If a given device has both + device-tree and ACPI support, it is presently (October 2020) + recommended to boot using the device-tree. + config GICV3 bool "GICv3 driver" depends on ARM_64 && !NEW_VGIC
xen/arch/arm/acpi/boot.c +9 -0
@@ -254,6 +254,15 @@ int __init acpi_boot_table_init(void) dt_scan_depth1_nodes, NULL) ) goto disable; +printk ("\n" +"*************************************************************************\n" +"* WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING *\n" +"* *\n" +"* Xen-ARM ACPI support is EXPERIMENTAL. It is presently (October 2020) *\n" +"* recommended you boot your system in device-tree mode if you can. *\n" +"*************************************************************************\n" +"\n" ); +/* * ACPI is disabled at this point. Enable it in order to parse * the ACPI tables.
Signature · 8 lines
2.20.1 -- (\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \BS ( | [email protected] PGP 87145445 | ) / \_CS\ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ 8A19\___\_|_/58D2 7E3D DDF4 7BA6 <-PGP-> 41D1 B375 37D0 8714\_|_/___/5445
Re: [PATCH] xen/arm: ACPI: Remove EXPERT dependancy, default on for ARM64
xen/arch/arm/Kconfig menu "Architecture Features" · line 37
Quoted 10 lines · Elliott Mitchell
xen/arch/arm/Kconfig +2
-1@@ -32,13 +32,18 @@ menu "Architecture Features" source "arch/Kconfig" config ACPI - bool "ACPI (Advanced Configuration and Power Interface) Support" if EXPERT + bool "ACPI (Advanced Configuration and Power Interface) Support" depends on ARM_64 + default y if ARM_64
The "if" is pointless with the "depends on".
Quoted 5 lines
xen/arch/arm/acpi/boot.c +0
-0@@ -254,6 +254,15 @@ int __init acpi_boot_table_init(void) dt_scan_depth1_nodes, NULL) ) goto disable;
xen/arch/arm/acpi/boot.c int __init acpi_boot_table_init(void) · line 265
Quoted 10 lines
xen/arch/arm/acpi/boot.c +9
-0+printk ("\n" +"*************************************************************************\n" +"* WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING *\n" +"* *\n" +"* Xen-ARM ACPI support is EXPERIMENTAL. It is presently (October 2020) *\n" +"* recommended you boot your system in device-tree mode if you can. *\n" +"*************************************************************************\n" +"\n" ); +
We have an abstraction for such warnings, causing them to appear later in the boot process and then consistently all in one place (both increasing, as we believe, the chances of being noticed):warning_add(). There's a delay accompanied with this, so I think you will want to also have a command line option allowing to silence this warning. "acpi=on" or "acpi=force", as available on x86 and (possibly wrongly right now) not documented as x86-specific, may be (re-)usable, i.e. to avoid having to introduce some entirely new option.
Also a few formal nits: The subject tag should have been [PATCH v2], there should have been a short revision log outside of the commit message area, and new patch versione would better start their own new threads than being in-reply-to the earlier version's one.
Jan
Re: [PATCH] xen/arm: ACPI: Remove EXPERT dependancy, default on for ARM64
In reply to Elliott MitchellQuoted 4 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI potentially taking over. Some common devices may require ACPI table support to boot.
Let's not make guesses on the direction of the industry in a commit message :-)
The following would suffice:
Some common ARM64 devices require ACPI to boot (no device tree is available).
Quoted 22 lines
For devices which can boot in either mode, continue defaulting to device-tree. Add warnings about using ACPI advising users of present situation.
Signed-off-by Elliott Mitchell <[email protected]>---
Okay, hopefully this is okay. Warning inKconfig, warning on boot. Perhaps "default y ifARM_64" is redundant, yet if someone tries to make it possible to boot aarch32 on a ACPI machine...I also want a date in the message. Theory is this won't be there forever, so a date is essential.
--- xen/arch/arm/Kconfig | 7 ++++++- xen/arch/arm/acpi/boot.c | 9 +++++++++ 2 files changed, 15 insertions(+), 1 deletion(-)xen/arch/arm/Kconfig +0
-0
xen/arch/arm/Kconfig menu "Architecture Features" · line 37
Quoted 7 lines
xen/arch/arm/Kconfig +2
-1@@ -32,13 +32,18 @@ menu "Architecture Features" source "arch/Kconfig" config ACPI - bool "ACPI (Advanced Configuration and Power Interface) Support" if EXPERT + bool "ACPI (Advanced Configuration and Power Interface) Support" depends on ARM_64 + default y if ARM_64
I am not so sure about the "default y" for the reason that the option is not technically "supported", so it is probably best to take the default line out. Otherwise we end up with a default "unsupported" kconfig which is not great.
xen/arch/arm/Kconfig menu "Architecture Features" · line 45
Quoted 8 lines
xen/arch/arm/Kconfig +3
-0---help--- Advanced Configuration and Power Interface (ACPI) support for Xen is an alternative to device tree on ARM64. + Note this is presently EXPERIMENTAL. If a given device has both + device-tree and ACPI support, it is presently (October 2020) + recommended to boot using the device-tree.
Please remove the date from the message. We'll update as needed in the future. The following works:
Note this is presently EXPERIMENTAL. If a given device has both device-tree and ACPI support, it is recommended to boot using the device-tree.
Quoted 9 lines
xen/arch/arm/Kconfig +0
-0config GICV3 bool "GICv3 driver" depends on ARM_64 && !NEW_VGICxen/arch/arm/acpi/boot.c +0
-0@@ -254,6 +254,15 @@ int __init acpi_boot_table_init(void) dt_scan_depth1_nodes, NULL) )
xen/arch/arm/acpi/boot.c int __init acpi_boot_table_init(void) · line 264
Quoted 10 lines
xen/arch/arm/acpi/boot.c +8
-0goto disable; +printk ("\n" +"*************************************************************************\n" +"* WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING *\n" +"* *\n" +"* Xen-ARM ACPI support is EXPERIMENTAL. It is presently (October 2020) *\n" +"* recommended you boot your system in device-tree mode if you can. *\n" +"*************************************************************************\n" +"\n" );
Please use warning_add and remove the date from the message.
Quoted 14 lines
xen/arch/arm/acpi/boot.c +0
-0/* * ACPI is disabled at this point. Enable it in order to parse * the ACPI tables. Signature · 8 lines
2.20.1 -- (\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \BS ( | [email protected] PGP 87145445 | ) / \_CS\ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ 8A19\___\_|_/58D2 7E3D DDF4 7BA6 <-PGP-> 41D1 B375 37D0 8714\_|_/___/5445
Hi Stefano,
Quoted 9 lines · Stefano Stabellini
Quoted 8 lines · Julien Grall
Quoted 6 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI potentially taking over. Some common devices may need ACPI table support.
Presently I think it is worth removing the dependancy on
CONFIG_EXPERT.
Quoted 7 lines
Quoted 2 lines
The idea behind EXPERT is to gate any feature that is not considered to be stable/complete enough to be used in production.
Yes, and from that point of view I don't think we want to remove EXPERT from ACPI yet. However, the idea of hiding things behind EXPERT works very well for new esoteric features, something like memory introspection or memory overcommit.
Memaccess is not very new ;).
Quoted 2 lines
It does not work well for things that are actually required to boot on the platform.
I am not sure where is the problem. It is easy to select EXPERT from the menuconfig. It also hints the user that the feature may not fully work.
Quoted 3 lines
Typically ACPI systems don't come with device tree at all (RPi4 being an exception), so users don't really have much of a choice in the matter.
And they typically have IOMMUs.
Quoted 10 lines
From that point of view, it would be better to remove EXPERT from ACPI,
maybe even build ACPI by default, *but* to add a warning at boot saying something like:"ACPI support is experimental. Boot using Device Tree if you can."
That would better convey the risks of using ACPI, while at the same time making it a bit easier for users to boot on their ACPI-only platforms.
Right, I agree that this make easier for users to boot Xen on ACPI-only platform. However, based on above, it is easy enough for a developper to rebuild Xen with ACPI and EXPERT enabled.
So what sort of users are you targeting?
I am sort of okay to remove EXPERT. But I still think building ACPI by default is still wrong because our default .config is meant to be (security) supported. I don't think ACPI can earn this qualification today.
In order to remove EXPERT, there are a few things to needs to be done (or checked):
1) SUPPORT.MD has a statement about ACPI on Arm
2) DT is favored over ACPI if the two firmware tables are present.
Quoted 11 lines
Quoted 7 lines
I don't consider the ACPI complete because the parsing of the IORT (used to discover SMMU and GICv3 ITS) is not there yet.
I vaguely remember some issues on system using SMMU (e.g. Thunder-X) because Dom0 will try to use the IOMMU and this would break PV drivers.
I am not sure why Dom0 using the IOMMU would break PV drivers? Is it because the pagetable is not properly updated when mapping foreign pages?
IIRC, yes. This would need to be tested again.
Cheers,
Signature · 1 line
Julien Grall
Quoted 43 lines · Julien Grall
Hi Stefano,
Quoted 10 lines · Stefano Stabellini
Quoted 9 lines · Julien Grall
Quoted 7 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI
potentially taking over. Some common devices may need ACPI table support.Presently I think it is worth removing the dependancy on
CONFIG_EXPERT.Quoted 7 lines
Quoted 2 lines
The idea behind EXPERT is to gate any feature that is not considered to be stable/complete enough to be used in production.
Yes, and from that point of view I don't think we want to remove EXPERT from ACPI yet. However, the idea of hiding things behind EXPERT works very well for new esoteric features, something like memory introspection or memory overcommit.
Memaccess is not very new ;).
Quoted 2 lines
It does not work well for things that are actually required to boot on the platform.
I am not sure where is the problem. It is easy to select EXPERT from the menuconfig. It also hints the user that the feature may not fully work.
Quoted 3 lines
Typically ACPI systems don't come with device tree at all (RPi4 being an exception), so users don't really have much of a choice in the matter.
And they typically have IOMMUs.
Quoted 8 lines
From that point of view, it would be better to remove EXPERT from ACPI,
maybe even build ACPI by default, *but* to add a warning at boot saying something like:"ACPI support is experimental. Boot using Device Tree if you can."
Quoted 8 lines
Quoted 2 lines
That would better convey the risks of using ACPI, while at the same time making it a bit easier for users to boot on their ACPI-only platforms.
Right, I agree that this make easier for users to boot Xen on ACPI-only platform. However, based on above, it is easy enough for a developper to rebuild Xen with ACPI and EXPERT enabled.
So what sort of users are you targeting?
Somebody trying Xen for the first time, they might know how to build it but they might not know that ACPI is not available by default, and they might not know that they need to enable EXPERT in order to get the ACPI option in the menu. It is easy to do once you know it is there, otherwise one might not know where to look in the menu.
Quoted 1 line
I am sort of okay to remove EXPERT.
OK. This would help (even without building it by default) because as you go and look at the menu the first time, you'll find ACPI among the options right away.
Quoted 3 lines
But I still think building ACPI by default
is still wrong because our default .config is meant to be (security) supported. I don't think ACPI can earn this qualification today.
Certainly we don't want to imply ACPI is security supported. I was looking at SUPPORT.md and it is only says:
"""
EXPERT and DEBUG Kconfig options are not security supported. Other Kconfig options are supported, if the related features are marked as supported in this document.
"""
So technically we could enable ACPI in the build by default as ACPI for ARM is marked as experimental. However, I can see that it is not a great idea to enable by default an unsupported option in the kconfig, so from that point of view it might be best to leave ACPI disabled by default. Probably the best compromise at this time.
Quoted 3 lines
In order to remove EXPERT, there are a few things to needs to be done (or checked):
1) SUPPORT.MD has a statement about ACPI on Arm
### Host ACPI (via Domain 0)
Status, x86 PV: Supported
Status, ARM: Experimental
Quoted 1 line
2) DT is favored over ACPI if the two firmware tables are present.
Good idea. xen/arch/arm/acpi/boot.c:acpi_boot_table_init has:
/*
* Enable ACPI instead of device tree unless
* - ACPI has been disabled explicitly (acpi=off), or
* - the device tree is not empty (it has more than just a /chosen node)
* and ACPI has not been force enabled (acpi=force)
*/
if ( param_acpi_off)
goto disable;
if ( !param_acpi_force &&
device_tree_for_each_node(device_tree_flattened, 0,
dt_scan_depth1_nodes, NULL) )
goto disable;
We should be fine.
Hi Stefano,
Quoted 51 lines · Stefano Stabellini
Quoted 50 lines · Julien Grall
Hi Stefano,
Quoted 10 lines · Stefano Stabellini
Quoted 9 lines · Julien Grall
Quoted 7 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI
potentially taking over. Some common devices may need ACPI table support.Presently I think it is worth removing the dependancy on
CONFIG_EXPERT.Quoted 7 lines
Quoted 2 lines
The idea behind EXPERT is to gate any feature that is not considered to be stable/complete enough to be used in production.
Yes, and from that point of view I don't think we want to remove EXPERT from ACPI yet. However, the idea of hiding things behind EXPERT works very well for new esoteric features, something like memory introspection or memory overcommit.
Memaccess is not very new ;).
Quoted 2 lines
It does not work well for things that are actually required to boot on the platform.
I am not sure where is the problem. It is easy to select EXPERT from the menuconfig. It also hints the user that the feature may not fully work.
Quoted 3 lines
Typically ACPI systems don't come with device tree at all (RPi4 being an exception), so users don't really have much of a choice in the matter.
And they typically have IOMMUs.
Quoted 10 lines
From that point of view, it would be better to remove EXPERT from ACPI,
maybe even build ACPI by default, *but* to add a warning at boot saying something like:"ACPI support is experimental. Boot using Device Tree if you can."
That would better convey the risks of using ACPI, while at the same time making it a bit easier for users to boot on their ACPI-only platforms.
Right, I agree that this make easier for users to boot Xen on ACPI-only platform. However, based on above, it is easy enough for a developper to rebuild Xen with ACPI and EXPERT enabled.
Quoted 7 lines
Quoted 1 line
So what sort of users are you targeting?
Somebody trying Xen for the first time, they might know how to build it but they might not know that ACPI is not available by default, and they might not know that they need to enable EXPERT in order to get the ACPI option in the menu. It is easy to do once you know it is there, otherwise one might not know where to look in the menu.
Right, EXPERT can now be enabled using Kconfig. So it is not very different from an option Foo has been hidden because the dependency Bar has not been selected.
It should be easy enough (if it is not we should fix it) to figure out the dependency when searching the option via menuconfig.
Quoted 7 lines
Quoted 3 lines
I am sort of okay to remove EXPERT.
OK. This would help (even without building it by default) because as you go and look at the menu the first time, you'll find ACPI among the options right away.
To be honest, this step is probably the easiest in the full process to get Xen build and booted on Arm.
I briefly looked at Elliot's v2, and I can't keep thinking that we are trying to re-invent EXPERT for ACPI because we think the feature is *more* important than any other feature gated by EXPERT.
In fact, all the features behind EXPERT are important. But they have been gated by EXPERT because they are not mature enough.
We already moved EXPERT from a command line option to a menuconfig option. So it should be easy enough to enable it now. If it still not the case, then we should improve it.
But I don't think ACPI is mature enough to deserve a different treatment. It would be more useful to get to the stage where ACPI can work without any crash/issue first.
Quoted 15 lines
Quoted 5 lines
But I still think building ACPI by default
is still wrong because our default .config is meant to be (security) supported. I don't think ACPI can earn this qualification today.Certainly we don't want to imply ACPI is security supported. I was looking at SUPPORT.md and it is only says:
"""
EXPERT and DEBUGKconfigoptions are not security supported. OtherKconfigoptions are supported, if the related features are marked as supported in this document.
"""
Quoted 5 lines
So technically we could enable ACPI in the build by default as ACPI for ARM is marked as experimental. However, I can see that it is not a great idea to enable by default an unsupported option in the kconfig, so from that point of view it might be best to leave ACPI disabled by default. Probably the best compromise at this time.
From my understanding, the goal of EXPERT was to gate such features. With your suggestion, it is not clear to me what's the difference between "experimental" and option gated by "EXPERT".
Do you mind clarifying?
Cheers,
Signature · 1 line
Julien Grall
Quoted 66 lines · Julien Grall
Hi Stefano,
Quoted 55 lines · Stefano Stabellini
Quoted 54 lines · Julien Grall
Hi Stefano,
Quoted 12 lines · Stefano Stabellini
Quoted 11 lines · Julien Grall
Quoted 9 lines · Elliott Mitchell
Linux requires UEFI support to be enabled on ARM64 devices. While many
ARM64 devices lack ACPI, the writing seems to be on the wall of UEFI/ACPI
potentially taking over. Some common devices may need ACPI table support.Presently I think it is worth removing the dependancy on
CONFIG_EXPERT.Quoted 8 lines
Quoted 3 lines
The idea behind EXPERT is to gate any feature that is not considered to be
stable/complete enough to be used in production.Yes, and from that point of view I don't think we want to remove EXPERT from ACPI yet. However, the idea of hiding things behind EXPERT works very well for new esoteric features, something like memory introspection or memory overcommit.
Memaccess is not very new ;).
Quoted 2 lines
It does not work well for things that are actually required to boot on the platform.
I am not sure where is the problem. It is easy to select EXPERT from the menuconfig. It also hints the user that the feature may not fully work.
Quoted 3 lines
Typically ACPI systems don't come with device tree at all (RPi4 being an exception), so users don't really have much of a choice in the matter.
And they typically have IOMMUs.
Quoted 11 lines
From that point of view, it would be better to remove EXPERT from
ACPI,
maybe even build ACPI by default, *but* to add a warning at boot saying something like:"ACPI support is experimental. Boot using Device Tree if you can."
That would better convey the risks of using ACPI, while at the same time making it a bit easier for users to boot on their ACPI-only platforms.
Right, I agree that this make easier for users to boot Xen on ACPI-only platform. However, based on above, it is easy enough for a developper to rebuild Xen with ACPI and EXPERT enabled.
Quoted 7 lines
Quoted 1 line
So what sort of users are you targeting?
Somebody trying Xen for the first time, they might know how to build it but they might not know that ACPI is not available by default, and they might not know that they need to enable EXPERT in order to get the ACPI option in the menu. It is easy to do once you know it is there, otherwise one might not know where to look in the menu.
Quoted 6 lines
Right, EXPERT can now be enabled using
Kconfig. So it is not very different from an option Foo has been hidden because the dependency Bar has not been selected.It should be easy enough (if it is not we should fix it) to figure out the dependency when searching the option via menuconfig.
I do make menuconfig and there is no ACPI option. How do I even know that ACPI has a kconfig option to enable? I'd assume that ACPI is always enabled in the kconfig unless told otherwise.
But let's say that you know that you need to look for ACPI. I'd use the search function, and it tells me:
Symbol: ACPI [=n] │
Type : bool │
Prompt: ACPI (Advanced Configuration and Power Interface) Support │
Location: │
(1) -> Architecture Features │
Defined at arch/arm/Kconfig:34 │
Depends on: ARM_64 [=y]I go and look "Architecture Features" as told, but it is not there. How do I need that I need to enable "Configure standard Xen features (expert users)" to get that option?
Quoted 9 lines
Quoted 5 lines
Quoted 1 line
I am sort of okay to remove EXPERT.
OK. This would help (even without building it by default) because as you go and look at the menu the first time, you'll find ACPI among the options right away.
To be honest, this step is probably the easiest in the full process to get Xen build and booted on Arm.
Quoted 6 lines
I briefly looked at Elliot's v2, and I can't keep thinking that we are trying to re-invent EXPERT for ACPI because we think the feature is *more* important than any other feature gated by EXPERT.
In fact, all the features behind EXPERT are important. But they have been gated by EXPERT because they are not mature enough.
It is not as much a matter of "importance" but a matter of required for booting. I don't think that EXPERT is really a good tool for gating things that are required for booting. If we had something else (not ACPI) that is required for booting and marked as EXPERT, I'd say the same thing. The only other thing that might qualify is ITS support.
Quoted 5 lines
We already moved EXPERT from a command line option to a menuconfig option. So it should be easy enough to enable it now. If it still not the case, then we should improve it.
But I don't think ACPI is mature enough to deserve a different treatment.
I fully agree ACPI is not mature.
Quoted 2 lines
It would be more useful to get to the stage where ACPI can work without any crash/issue first.
Yes indeed. I don't have any stake in this, given that none of my systems have ACPI, so I'd better shut up maybe :-)
Quoted 18 lines
Quoted 13 lines
Quoted 3 lines
But I still think building ACPI by default
is still wrong because our default .config is meant to be (security) supported. I don't think ACPI can earn this qualification today.Certainly we don't want to imply ACPI is security supported. I was looking at SUPPORT.md and it is only says:
"""
EXPERT and DEBUGKconfigoptions are not security supported. OtherKconfigoptions are supported, if the related features are marked as supported in this document.
"""Quoted 5 lines
So technically we could enable ACPI in the build by default as ACPI for ARM is marked as experimental. However, I can see that it is not a great idea to enable by default an unsupported option in the kconfig, so from that point of view it might be best to leave ACPI disabled by default. Probably the best compromise at this time.
Quoted 5 lines
From my understanding, the goal of EXPERT was to gate such features. With your suggestion, it is not clear to me what's the difference between "experimental" and option gated by "EXPERT".
Do you mind clarifying?
Ah! That's a good question actually. Is the expectation and "experimental" in SUPPORT.md and EXPERT in Kconfig are pretty much the same thing? I didn't think so. Let's have a look. SUPPORT.md says:
### KCONFIG Expert
Status: Experimental
And EXPERT says:
config EXPERT
bool "Configure standard Xen features (expert users)" help
This option allows certain base Xen options and settings to be disabled or tweaked. This is for specialized environments which can tolerate a "non-standard" Xen.
Only use this if you really know what you are doing. Xen binaries built with this option enabled are not security supported.
It seems that they are not the same: EXPERT is a *subset* of Experimental for certain Xen options "for specialized environments" and "expert users, which I think makes sense. ACPI is a good example of something "experimental" but not for specialized environments.
Quoted 9 lines · Julien Grall
Quoted 7 lines · Stefano Stabellini
Quoted 1 line · Julien Grall
I am sort of okay to remove EXPERT.
OK. This would help (even without building it by default) because as you go and look at the menu the first time, you'll find ACPI among the options right away.
Quoted 6 lines
To be honest, this step is probably the easiest in the full process to get Xen build and booted on Arm.
I briefly looked at Elliot's v2, and I can't keep thinking that we are trying to re-invent EXPERT for ACPI because we think the feature is *more* important than any other feature gated by EXPERT.
Yet might that statement in fact be true?
Most of the features currently controlled by CONFIG_EXPERT are relatively small tweaks which enable less often used features. Some of those are very high value in certain environments, but they're unimportant in common environments. Changing the scheduler might get you an extra 10-50% performance improvement in a special environment.
ACPI support isn't like this. I'm unaware what Masami Hiramatsu's system does if a CONFIG_ACPI=n Xen build is tried. I haven't actually tried such a build on mine, but from the code it looks like Xen would panic if built that way. No output of any sort. Simply panic with the device appearing to go inert.
Quoted 6 lines
In fact, all the features behind EXPERT are important. But they have been gated by EXPERT because they are not mature enough.
But I don't think ACPI is mature enough to deserve a different treatment. It would be more useful to get to the stage where ACPI can work without any crash/issue first.
The difference is the severity of failure if the option is off, but needs to be on. Most CONFIG_EXPERT options will merely be performance reductions or security situations unacceptable to some users. CONFIG_ACPI=n on the wrong system could be a panic with *no* output.
Tainting sounds reasonable. Messages in dmesg make sense. A message plus 10 second pause on boot seems reasonable. I think if CONFIG_ACPI is off, there should be code to try to detect ACPI and emit a warning if anything is detected.
Signature · 4 lines
(\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \BS ( | [email protected] PGP 87145445 | ) / \_CS\ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ 8A19\___\_|_/58D2 7E3D DDF4 7BA6 <-PGP-> 41D1 B375 37D0 8714\_|_/___/5445
Hi Elliott,
Quoted 10 lines · Elliott Mitchell
Quoted 9 lines · Julien Grall
Quoted 7 lines · Stefano Stabellini
Quoted 1 line · Julien Grall
I am sort of okay to remove EXPERT.
OK. This would help (even without building it by default) because as you go and look at the menu the first time, you'll find ACPI among the options right away.
Quoted 8 lines
Quoted 6 lines
To be honest, this step is probably the easiest in the full process to get Xen build and booted on Arm.
I briefly looked at Elliot's v2, and I can't keep thinking that we are trying to re-invent EXPERT for ACPI because we think the feature is *more* important than any other feature gated by EXPERT.
Yet might that statement in fact be true?
Everyone has a different opinion on what's important or not. I am sure we can spend hours bikeshedding on that...
Quoted 7 lines
Most of the features currently controlled by
CONFIG_EXPERTare relatively small tweaks which enable less often used features. Some of those are very high value in certain environments, but they're unimportant in common environments. Changing the scheduler might get you an extra 10-50% performance improvement in a special environment.
Quoted 5 lines
ACPI support isn't like this. I'm unaware what Masami Hiramatsu's system does if a
CONFIG_ACPI=nXen build is tried. I haven't actually tried such a build on mine, but from the code it looks like Xen would panic if built that way. No output of any sort. Simply panic with the device appearing to go inert.
There will always be cases where the console is not working:
1) There is no driver in Xen
2) There is no SPCR table present
I think you are in the second situation and you had to enable earlyprintk. Is that correct?
Quoted 12 lines
Quoted 6 lines
In fact, all the features behind EXPERT are important. But they have been gated by EXPERT because they are not mature enough.
But I don't think ACPI is mature enough to deserve a different treatment. It would be more useful to get to the stage where ACPI can work without any crash/issue first.
The difference is the severity of failure if the option is off, but needs to be on. Most
CONFIG_EXPERToptions will merely be performance reductions or security situations unacceptable to some users.CONFIG_ACPI=non the wrong system could be a panic with *no* output. >
Quoted 4 lines
Tainting sounds reasonable. Messages in
dmesgmake sense. A message plus 10 second pause on boot seems reasonable. I think ifCONFIG_ACPIis off, there should be code to try to detect ACPI and emit a warning if anything is detected.
All of this would be moot if we focus on getting ACPI (or just a subset) working on a few platforms (e.g. RPI, Thunder-X).
I don't think we are too far from this acheviement. This would be IMHO be enough to move ACPI one way up in the support "ladder". We might even be able to do this for 4.15.
Cheers,
Signature · 1 line
Julien Grall