Fixes for deferred_probe_timeout cleanup
Subsystems (6)
v3 cover Fixes for deferred_probe_timeout cleanup
Just wanted to resubmit these three fixes for the deferred_probe_timeout cleanup that landed in the v5.7-rc1 merge window.
The first resets the default timeout value back to zero so we have no behavioral change from 5.6. This avoids regressions on boards that have "optional links" in their device tree.
The second changes the code to use dev_warn() instead of dev_WARN() to address complaints about unnecessary backtraces in the boot log.
The last fixes an issue discovered by Yoshihiro Shimoda and Geert Uytterhoeven, where if a timeout was set, things like NFS root might fail due to wait_for_device_probe() not blocking until the timeout expires.
thanks
-john
New in v3:
* Just included the previously posted dev_warn() fix into the series, so they didn't collide.
John Stultz (3):
driver core: Revert default driver_deferred_probe_timeout value to 0 driver core: Use dev_warn() instead of dev_WARN() for
deferred_probe_timeout warnings
driver core: Ensure wait_for_device_probe() waits until the
deferred_probe_timeout fires
drivers/base/dd.c | 20 ++++++++------------ 1 file changed, 8 insertions(+), 12 deletions(-)
Signature · 1 line
2.17.1
PATCH v3 1/3 driver core: Revert default driver_deferred_probe_timeout value to 0
This patch addresses a regression in 5.7-rc1+
In commit c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic"), we both cleaned up the logic and also set the default driver_deferred_probe_timeout value to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland after initcalls_done is set.
However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the ip_auto_config() had already failed, resulting in NFS to fail. This was due to ip_auto_config() calling wait_for_device_probe() which doesn't wait for the driver_deferred_probe_timeout to fire.
Fixing that issue is possible, but could also introduce 30 second delays in bootups for users who don't have any missing dependencies, which is not ideal.
So I think the best solution to avoid any regressions is to revert back to a default timeout value of zero, and allow systems that need to utilize the timeout in order for userland to load any modules that supply misisng dependencies in the dts to specify the timeout length via the exiting documented boot argument.
Thanks to Geert for chasing down that ip_auto_config was why NFS was failing in this case!
--- drivers/base/dd.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-)
drivers/base/dd.c +2 -11
@@ -224,16 +224,7 @@ static int deferred_devs_show(struct seq_file *s, void *data) }DEFINE_SHOW_ATTRIBUTE (deferred_devs); -#ifdef CONFIG_MODULES -/* -* In the case of modules, set the default probe timeout to -* 30 seconds to give userland some time to load needed modules -*/ -int driver_deferred_probe_timeout =30 ; -#else -/* In the case of !modules, no probe timeout needed */ -int driver_deferred_probe_timeout =-1 ; -#endif +int driver_deferred_probe_timeout;EXPORT_SYMBOL_GPL (driver_deferred_probe_timeout);static int __init deferred_probe_timeout_setup (char *str)
@@ -266,7 +257,7 @@ int driver_deferred_probe_check_state(struct device *dev)return -ENODEV ; } -if (!driver_deferred_probe_timeout) { +if (!driver_deferred_probe_timeout && initcalls_done) {dev_WARN (dev,"deferred probe timeout, ignoring dependency" );return -ETIMEDOUT ; }
Signature · 1 line
2.17.1
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
Hi John,
Quoted 50 lines · John Stultz
This patch addresses a regression in 5.7-rc1+
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic"), we both cleaned up the logic and also set the defaultdriver_deferred_probe_timeoutvalue to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland afterinitcalls_doneis set.However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the
ip_auto_config()had already failed, resulting in NFS to fail. This was due toip_auto_config()callingwait_for_device_probe()which doesn't wait for thedriver_deferred_probe_timeoutto fire.Fixing that issue is possible, but could also introduce 30 second delays in bootups for users who don't have any missing dependencies, which is not ideal.
So I think the best solution to avoid any regressions is to revert back to a default timeout value of zero, and allow systems that need to utilize the timeout in order for userland to load any modules that supply misisng dependencies in the dts to specify the timeout length via the exiting documented boot argument.
Thanks to Geert for chasing down that
ip_auto_configwas why NFS was failing in this case!Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>
Quoted 8 lines
Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Reported-by Yoshihiro Shimoda <[email protected]>Tested-by Yoshihiro Shimoda <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>
Looks like you lost my
on v2.
Gr{oetje,eeting}s,
Geert
Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected] In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
In reply to John StultzHi John,
Quoted 50 lines · John Stultz
This patch addresses a regression in 5.7-rc1+
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic"), we both cleaned up the logic and also set the defaultdriver_deferred_probe_timeoutvalue to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland afterinitcalls_doneis set.However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the
ip_auto_config()had already failed, resulting in NFS to fail. This was due toip_auto_config()callingwait_for_device_probe()which doesn't wait for thedriver_deferred_probe_timeoutto fire.Fixing that issue is possible, but could also introduce 30 second delays in bootups for users who don't have any missing dependencies, which is not ideal.
So I think the best solution to avoid any regressions is to revert back to a default timeout value of zero, and allow systems that need to utilize the timeout in order for userland to load any modules that supply misisng dependencies in the dts to specify the timeout length via the exiting documented boot argument.
Thanks to Geert for chasing down that
ip_auto_configwas why NFS was failing in this case!Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>
Quoted 8 lines
Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Reported-by Yoshihiro Shimoda <[email protected]>Tested-by Yoshihiro Shimoda <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>
Please also revert dca0b44957e5 "regulator: Use driver_deferred_probe_timeout for regulator_init_complete_work" then, because now with the default 0 timeout some regulators gets disabled during boot, before their supplies gets instantiated.
This patch broke booting of Samsung Exynos5800-based Peach-Pi Chromeboot with the default multi_v7_defconfig.
Quoted 27 lines
--- drivers/base/dd.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-)drivers/base/dd.c +1
-10@@ -224,16 +224,7 @@ static int deferred_devs_show(struct seq_file *s, void *data) }DEFINE_SHOW_ATTRIBUTE (deferred_devs); -#ifdef CONFIG_MODULES -/* -* In the case of modules, set the default probe timeout to -* 30 seconds to give userland some time to load needed modules -*/ -int driver_deferred_probe_timeout =30 ; -#else -/* In the case of !modules, no probe timeout needed */ -int driver_deferred_probe_timeout =-1 ; -#endif +int driver_deferred_probe_timeout;EXPORT_SYMBOL_GPL (driver_deferred_probe_timeout);static int __init deferred_probe_timeout_setup (char *str)
drivers/base/dd.c int driver_deferred_probe_check_state(struct device *dev) · line 263
Quoted 8 lines
drivers/base/dd.c +1
-1@@ -266,7 +257,7 @@ int driver_deferred_probe_check_state(struct device *dev)return -ENODEV ; } -if (!driver_deferred_probe_timeout) { +if (!driver_deferred_probe_timeout && initcalls_done) {dev_WARN (dev,"deferred probe timeout, ignoring dependency" );return -ETIMEDOUT ; }
Best regards
Signature · 2 lines
Marek Szyprowski, PhD Samsung R&D Institute Poland
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
Quoted 9 lines · Marek Szyprowski
Quoted 2 lines · John Stultz
Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>Please also revert
dca0b44957e5"regulator: Usedriver_deferred_probe_timeoutforregulator_init_complete_work" then, because now with the default 0 timeout some regulators gets disabled during boot, before their supplies gets instantiated.
Yes, please - I requested this when the revert was originally proposed :(
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
Quoted 6 lines · Mark Brown
Quoted 4 lines · Marek Szyprowski
Quoted 2 lines · John Stultz
Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>
Quoted 6 lines
Quoted 4 lines
Please also revert
dca0b44957e5"regulator: Usedriver_deferred_probe_timeoutforregulator_init_complete_work" then, because now with the default 0 timeout some regulators gets disabled during boot, before their supplies gets instantiated.Yes, please - I requested this when the revert was originally proposed :(
Oh, my apologies. I misunderstood what you were suggesting earlier. Sorry for being thick headed.
I'll spin up a revert here shortly.
thanks
-john
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
In reply to John StultzQuoted 58 lines · John Stultz
This patch addresses a regression in 5.7-rc1+
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic"), we both cleaned up the logic and also set the defaultdriver_deferred_probe_timeoutvalue to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland afterinitcalls_doneis set.However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the
ip_auto_config()had already failed, resulting in NFS to fail. This was due toip_auto_config()callingwait_for_device_probe()which doesn't wait for thedriver_deferred_probe_timeoutto fire.Fixing that issue is possible, but could also introduce 30 second delays in bootups for users who don't have any missing dependencies, which is not ideal.
So I think the best solution to avoid any regressions is to revert back to a default timeout value of zero, and allow systems that need to utilize the timeout in order for userland to load any modules that supply misisng dependencies in the dts to specify the timeout length via the exiting documented boot argument.
Thanks to Geert for chasing down that
ip_auto_configwas why NFS was failing in this case!Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Reported-by Yoshihiro Shimoda <[email protected]>Tested-by Yoshihiro Shimoda <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>
Quoted 3 lines
--- drivers/base/dd.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-)
Sorry for being a bit late to the party, but this breaks suspend/resume support on various Tegra devices. I've only noticed now because, well, suspend/resume have been broken for other reasons for a little while and it's taken us a bit to resolve those issues.
But now that those other issues have been fixed, I've started seeing an issue where after resume from suspend some of the I2C controllers are no longer working. The reason for this is that they share pins with DP AUX controllers via the pinctrl framework. The DP AUX driver registers as part of the DRM/KMS driver, which usually happens in userspace. Since the deferred probe timeout was set to 0 by default this no longer works because no pinctrl states are assigned to the I2C controller and therefore upon resume the pins cannot be configured for I2C operation.
I'm also somewhat confused by this patch and a few before because they claim that they restore previous default behaviour, but that's just not true. Originally when this timeout was introduced it was -1, which meant that there was no timeout at all and hence users had to opt-in if they wanted to use a deferred probe timeout.
But now after this series the default is for there to be a very short timeout, which in turn causes existing use-cases to potentially break. I'm also going to suggest here that in most cases a driver will require the resources that it asks for, so the case that Yoshihiro described and that this patch is meant to fix sounds to me like it's the odd one out rather than the other way around.
But I realize that that's not very constructive. So perhaps we can find some other way for drivers to advertise that their dependencies are optional? I came up with the below patch, which restores suspend/resume on Tegra and could be used in conjunction with a patch that opts into this behaviour for the problematic driver in Yoshihiro's case to make this again work for everyone.
--- >8 ---
From a95f8f41b8a32dee3434db4f0515af7376d1873a Mon Sep 17 00:00:00 2001
From: Thierry Reding <[email protected]>
Date: Thu, 6 Aug 2020 14:51:59 +0200
Subject: [PATCH] driver core: Do not ignore dependencies by default
Many drivers do require the resources that they ask for and timing out may not always be an option. While there is a way to allow probing to continue to be deferred for some time after the system has booted, the fact that this is controlled via a command-line parameter is undesired because it require manual intervention, whereas in can be avoid in the majority of cases.
Instead of requiring users to edit the kernel command-line, add a way for drivers to specify whether or not their dependencies are optional so that they can continue deferring probe indefinitely.
--- drivers/base/dd.c | 2 +- include/linux/device/driver.h | 1 + 2 files changed, 2 insertions(+), 1 deletion(-)
drivers/base/dd.c +1 -1
@@ -279,7 +279,7 @@ int driver_deferred_probe_check_state(struct device *dev)return -ENODEV ; } -if (!driver_deferred_probe_timeout && initcalls_done) { +if (dev->driver->ignore_dependencies && !driver_deferred_probe_timeout && initcalls_done) {dev_warn (dev,"deferred probe timeout, ignoring dependency\n" );return -ETIMEDOUT ; }
include/linux/device/driver.h +1 -0
@@ -100,6 +100,7 @@ struct device_driver {const char *mod_name;/* used for built-in modules */ bool suppress_bind_attrs;/* disables bind/unbind via sysfs */ +bool ignore_dependencies;/* ignores dependencies */ enum probe_type probe_type;const struct of_device_id *of_match_table;
Signature · 14 lines
2.27.0 --- >8 --- Although, thinking about it a bit more it sounds to me like an even better approach would be to make this part of the API where a resource is requested. There are in fact already APIs that can request optional resources (such as regulator_get_optional()), so I think it would make more sense for any driver that can live without a resource to request it with an optional flag, which in turn could then trigger this code path for the deferred probe timeout. For anyone that really needs the resources that they request, they really shouldn't have to jump through hoops to get there. Thierry
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
Quoted 69 lines · Thierry Reding
Quoted 58 lines · John Stultz
This patch addresses a regression in 5.7-rc1+
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic"), we both cleaned up the logic and also set the defaultdriver_deferred_probe_timeoutvalue to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland afterinitcalls_doneis set.However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the
ip_auto_config()had already failed, resulting in NFS to fail. This was due toip_auto_config()callingwait_for_device_probe()which doesn't wait for thedriver_deferred_probe_timeoutto fire.Fixing that issue is possible, but could also introduce 30 second delays in bootups for users who don't have any missing dependencies, which is not ideal.
So I think the best solution to avoid any regressions is to revert back to a default timeout value of zero, and allow systems that need to utilize the timeout in order for userland to load any modules that supply misisng dependencies in the dts to specify the timeout length via the exiting documented boot argument.
Thanks to Geert for chasing down that
ip_auto_configwas why NFS was failing in this case!Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Reported-by Yoshihiro Shimoda <[email protected]>Tested-by Yoshihiro Shimoda <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>Quoted 3 lines
--- drivers/base/dd.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-)Sorry for being a bit late to the party, but this breaks suspend/resume support on various Tegra devices. I've only noticed now because, well, suspend/resume have been broken for other reasons for a little while and it's taken us a bit to resolve those issues.
Quoted 8 lines
But now that those other issues have been fixed, I've started seeing an issue where after resume from suspend some of the I2C controllers are no longer working. The reason for this is that they share pins with DP AUX controllers via the pinctrl framework. The DP AUX driver registers as part of the DRM/KMS driver, which usually happens in userspace. Since the deferred probe timeout was set to 0 by default this no longer works because no pinctrl states are assigned to the I2C controller and therefore upon resume the pins cannot be configured for I2C operation.
Oof. My apologies!
Quoted 5 lines
I'm also somewhat confused by this patch and a few before because they claim that they restore previous default behaviour, but that's just not true. Originally when this timeout was introduced it was -1, which meant that there was no timeout at all and hence users had to opt-in if they wanted to use a deferred probe timeout.
I don't think that's quite true, since the point of my original changes were to avoid troubles I was seeing with drivers not loading because once the timeout fired after init, driver loading would fail with ENODEV instead of returning EPROBE_DEFER. The logic that existed was buggy so the timeout handling didn't really work (changing the boot argument wouldn't help, because after init the logic would return ENODEV before it checked the timeout value).
That said, looking at it now, I do realize the driver_deferred_probe_check_state_continue() logic in effect never returned ETIMEDOUT before was consolidated in the earlier changes, and now we've backed the default timeout to 0, old user (see bec6c0ecb243) will now get ETIMEDOUT where they wouldn't before.
So would the following fix it up for you? (sorry its whitespace corrupted)
drivers/pinctrl/devicetree.c +0 -1
@@ -129,9 +129,8 @@ static int dt_to_map_one_config(struct pinctrl *p,if (!np_pctldev ||of_node_is_root (np_pctldev)) {of_node_put (np_pctldev); ret =driver_deferred_probe_check_state (p->dev); -/* keep deferring if modules are enabled
unless we've timed out */
- if (IS_ENABLED(CONFIG_MODULES) && !allow_default &&
- (ret == -ENODEV))
+ /* keep deferring if modules are enabled */
+ if (IS_ENABLED(CONFIG_MODULES) &&
!allow_default && ret < 0)
ret = -EPROBE_DEFER;
return ret;
}(you could probably argue calling driver_deferred_probe_check_state checking ret at all is silly here, since EPROBE_DEFER is the only option you want)
Quoted 66 lines
But now after this series the default is for there to be a very short timeout, which in turn causes existing use-cases to potentially break. I'm also going to suggest here that in most cases a driver will require the resources that it asks for, so the case that Yoshihiro described and that this patch is meant to fix sounds to me like it's the odd one out rather than the other way around.
But I realize that that's not very constructive. So perhaps we can find some other way for drivers to advertise that their dependencies are optional? I came up with the below patch, which restores suspend/resume on Tegra and could be used in conjunction with a patch that opts into this behaviour for the problematic driver in Yoshihiro's case to make this again work for everyone.
--- >8 ---
Froma95f8f41b8a32dee3434db4f0515af7376d1873aMon Sep 17 00:00:00 2001
From: Thierry Reding <[email protected]>
Date: Thu, 6 Aug 2020 14:51:59 +0200
Subject: [PATCH] driver core: Do not ignore dependencies by defaultMany drivers do require the resources that they ask for and timing out may not always be an option. While there is a way to allow probing to continue to be deferred for some time after the system has booted, the fact that this is controlled via a command-line parameter is undesired because it require manual intervention, whereas in can be avoid in the majority of cases.
Instead of requiring users to edit the kernel command-line, add a way for drivers to specify whether or not their dependencies are optional so that they can continue deferring probe indefinitely.
Signed-off-by Thierry Reding <[email protected]>--- drivers/base/dd.c | 2 +- include/linux/device/driver.h | 1 + 2 files changed, 2 insertions(+), 1 deletion(-)drivers/base/dd.c +1
-1@@ -279,7 +279,7 @@ int driver_deferred_probe_check_state(struct device *dev)return -ENODEV ; } -if (!driver_deferred_probe_timeout && initcalls_done) { +if (dev->driver->ignore_dependencies && !driver_deferred_probe_timeout && initcalls_done) {dev_warn (dev,"deferred probe timeout, ignoring dependency\n" );return -ETIMEDOUT ; }include/linux/device/driver.h +1
-1@@ -100,6 +100,7 @@ struct device_driver {const char *mod_name;/* used for built-in modules */ bool suppress_bind_attrs;/* disables bind/unbind via sysfs */ +bool ignore_dependencies;/* ignores dependencies */ enum probe_type probe_type;const struct of_device_id *of_match_table; --2.27.0
--- >8 ---
Quoted 9 lines
Although, thinking about it a bit more it sounds to me like an even better approach would be to make this part of the API where a resource is requested. There are in fact already APIs that can request optional resources (such as
regulator_get_optional()), so I think it would make more sense for any driver that can live without a resource to request it with an optional flag, which in turn could then trigger this code path for the deferred probe timeout. For anyone that really needs the resources that they request, they really shouldn't have to jump through hoops to get there.
I do like these suggestions. I feel like the optional dt links handling is very opaque as there's no real way to tell if a dt link is truly optional or not. Having the timeout is a really subtle and implicit way of getting that optional link handling behavior.
thanks
-john
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
Quoted 118 lines · John Stultz
Quoted 69 lines · Thierry Reding
Quoted 58 lines · John Stultz
This patch addresses a regression in 5.7-rc1+
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic"), we both cleaned up the logic and also set the defaultdriver_deferred_probe_timeoutvalue to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland afterinitcalls_doneis set.However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the
ip_auto_config()had already failed, resulting in NFS to fail. This was due toip_auto_config()callingwait_for_device_probe()which doesn't wait for thedriver_deferred_probe_timeoutto fire.Fixing that issue is possible, but could also introduce 30 second delays in bootups for users who don't have any missing dependencies, which is not ideal.
So I think the best solution to avoid any regressions is to revert back to a default timeout value of zero, and allow systems that need to utilize the timeout in order for userland to load any modules that supply misisng dependencies in the dts to specify the timeout length via the exiting documented boot argument.
Thanks to Geert for chasing down that
ip_auto_configwas why NFS was failing in this case!Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Reported-by Yoshihiro Shimoda <[email protected]>Tested-by Yoshihiro Shimoda <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>Quoted 3 lines
--- drivers/base/dd.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-)Sorry for being a bit late to the party, but this breaks suspend/resume support on various Tegra devices. I've only noticed now because, well, suspend/resume have been broken for other reasons for a little while and it's taken us a bit to resolve those issues.
Quoted 8 lines
But now that those other issues have been fixed, I've started seeing an issue where after resume from suspend some of the I2C controllers are no longer working. The reason for this is that they share pins with DP AUX controllers via the pinctrl framework. The DP AUX driver registers as part of the DRM/KMS driver, which usually happens in userspace. Since the deferred probe timeout was set to 0 by default this no longer works because no pinctrl states are assigned to the I2C controller and therefore upon resume the pins cannot be configured for I2C operation.
Oof. My apologies!
Quoted 5 lines
I'm also somewhat confused by this patch and a few before because they claim that they restore previous default behaviour, but that's just not true. Originally when this timeout was introduced it was -1, which meant that there was no timeout at all and hence users had to opt-in if they wanted to use a deferred probe timeout.
I don't think that's quite true, since the point of my original changes were to avoid troubles I was seeing with drivers not loading because once the timeout fired after init, driver loading would fail with ENODEV instead of returning
EPROBE_DEFER. The logic that existed was buggy so the timeout handling didn't really work (changing the boot argument wouldn't help, because after init the logic would return ENODEV before it checked the timeout value).That said, looking at it now, I do realize the
driver_deferred_probe_check_state_continue()logic in effect never returned ETIMEDOUT before was consolidated in the earlier changes, and now we've backed the default timeout to 0, old user (seebec6c0ecb243) will now get ETIMEDOUT where they wouldn't before.So would the following fix it up for you? (sorry its whitespace corrupted)
drivers/pinctrl/devicetree.c +0
-1@@ -129,9 +129,8 @@ static int dt_to_map_one_config(struct pinctrl *p,if (!np_pctldev ||of_node_is_root (np_pctldev)) {of_node_put (np_pctldev); ret =driver_deferred_probe_check_state (p->dev); -/* keep deferring if modules are enabled unless we've timed out */
drivers/pinctrl/devicetree.c +2
-2-if (IS_ENABLED (CONFIG_MODULES ) && !allow_default && - (ret == -ENODEV )) +/* keep deferring if modules are enabled */ + if (IS_ENABLED(CONFIG_MODULES) &&!allow_default && ret < 0)
Quoted 7 lines
ret = -EPROBE_DEFER; return ret; }(you could probably argue calling
driver_deferred_probe_check_statechecking ret at all is silly here, sinceEPROBE_DEFERis the only option you want)
Just by looking at it I would've said, yes, this looks like it would fix it. However, upon giving this a try, I see the same pinmux issues as I was seeing before. I'm going to have to dig a little deeper to see if I can find out why that is, because it isn't obvious to me right away.
Quoted 67 lines
Quoted 67 lines
But now after this series the default is for there to be a very short timeout, which in turn causes existing use-cases to potentially break. I'm also going to suggest here that in most cases a driver will require the resources that it asks for, so the case that Yoshihiro described and that this patch is meant to fix sounds to me like it's the odd one out rather than the other way around.
But I realize that that's not very constructive. So perhaps we can find some other way for drivers to advertise that their dependencies are optional? I came up with the below patch, which restores suspend/resume on Tegra and could be used in conjunction with a patch that opts into this behaviour for the problematic driver in Yoshihiro's case to make this again work for everyone.
--- >8 ---
Froma95f8f41b8a32dee3434db4f0515af7376d1873aMon Sep 17 00:00:00 2001
From: Thierry Reding <[email protected]>
Date: Thu, 6 Aug 2020 14:51:59 +0200
Subject: [PATCH] driver core: Do not ignore dependencies by defaultMany drivers do require the resources that they ask for and timing out may not always be an option. While there is a way to allow probing to continue to be deferred for some time after the system has booted, the fact that this is controlled via a command-line parameter is undesired because it require manual intervention, whereas in can be avoid in the majority of cases.
Instead of requiring users to edit the kernel command-line, add a way for drivers to specify whether or not their dependencies are optional so that they can continue deferring probe indefinitely.
Signed-off-by Thierry Reding <[email protected]>--- drivers/base/dd.c | 2 +- include/linux/device/driver.h | 1 + 2 files changed, 2 insertions(+), 1 deletion(-)drivers/base/dd.c +1
-1@@ -279,7 +279,7 @@ int driver_deferred_probe_check_state(struct device *dev)return -ENODEV ; } -if (!driver_deferred_probe_timeout && initcalls_done) { +if (dev->driver->ignore_dependencies && !driver_deferred_probe_timeout && initcalls_done) {dev_warn (dev,"deferred probe timeout, ignoring dependency\n" );return -ETIMEDOUT ; }include/linux/device/driver.h +1
-1@@ -100,6 +100,7 @@ struct device_driver {const char *mod_name;/* used for built-in modules */ bool suppress_bind_attrs;/* disables bind/unbind via sysfs */ +bool ignore_dependencies;/* ignores dependencies */ enum probe_type probe_type;const struct of_device_id *of_match_table; --2.27.0
--- >8 ---
Quoted 14 lines
Quoted 9 lines
Although, thinking about it a bit more it sounds to me like an even better approach would be to make this part of the API where a resource is requested. There are in fact already APIs that can request optional resources (such as
regulator_get_optional()), so I think it would make more sense for any driver that can live without a resource to request it with an optional flag, which in turn could then trigger this code path for the deferred probe timeout. For anyone that really needs the resources that they request, they really shouldn't have to jump through hoops to get there.I do like these suggestions. I feel like the optional dt links handling is very opaque as there's no real way to tell if a dt link is truly optional or not. Having the timeout is a really subtle and implicit way of getting that optional link handling behavior.
Agreed. Handling this implicitly is pretty wonky and as evidenced by these two cases bound to break one way or another.
Thierry
Re: [PATCH v3 1/3] driver core: Revert default driver_deferred_probe_timeout value to 0
Quoted 123 lines · Thierry Reding
Quoted 122 lines · John Stultz
Quoted 69 lines · Thierry Reding
Quoted 58 lines · John Stultz
This patch addresses a regression in 5.7-rc1+
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic"), we both cleaned up the logic and also set the defaultdriver_deferred_probe_timeoutvalue to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland afterinitcalls_doneis set.However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the
ip_auto_config()had already failed, resulting in NFS to fail. This was due toip_auto_config()callingwait_for_device_probe()which doesn't wait for thedriver_deferred_probe_timeoutto fire.Fixing that issue is possible, but could also introduce 30 second delays in bootups for users who don't have any missing dependencies, which is not ideal.
So I think the best solution to avoid any regressions is to revert back to a default timeout value of zero, and allow systems that need to utilize the timeout in order for userland to load any modules that supply misisng dependencies in the dts to specify the timeout length via the exiting documented boot argument.
Thanks to Geert for chasing down that
ip_auto_configwas why NFS was failing in this case!Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Reported-by Yoshihiro Shimoda <[email protected]>Tested-by Yoshihiro Shimoda <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>Quoted 3 lines
--- drivers/base/dd.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-)Sorry for being a bit late to the party, but this breaks suspend/resume support on various Tegra devices. I've only noticed now because, well, suspend/resume have been broken for other reasons for a little while and it's taken us a bit to resolve those issues.
Quoted 8 lines
But now that those other issues have been fixed, I've started seeing an issue where after resume from suspend some of the I2C controllers are no longer working. The reason for this is that they share pins with DP AUX controllers via the pinctrl framework. The DP AUX driver registers as part of the DRM/KMS driver, which usually happens in userspace. Since the deferred probe timeout was set to 0 by default this no longer works because no pinctrl states are assigned to the I2C controller and therefore upon resume the pins cannot be configured for I2C operation.
Oof. My apologies!
Quoted 5 lines
I'm also somewhat confused by this patch and a few before because they claim that they restore previous default behaviour, but that's just not true. Originally when this timeout was introduced it was -1, which meant that there was no timeout at all and hence users had to opt-in if they wanted to use a deferred probe timeout.
I don't think that's quite true, since the point of my original changes were to avoid troubles I was seeing with drivers not loading because once the timeout fired after init, driver loading would fail with ENODEV instead of returning
EPROBE_DEFER. The logic that existed was buggy so the timeout handling didn't really work (changing the boot argument wouldn't help, because after init the logic would return ENODEV before it checked the timeout value).That said, looking at it now, I do realize the
driver_deferred_probe_check_state_continue()logic in effect never returned ETIMEDOUT before was consolidated in the earlier changes, and now we've backed the default timeout to 0, old user (seebec6c0ecb243) will now get ETIMEDOUT where they wouldn't before.So would the following fix it up for you? (sorry its whitespace corrupted)
drivers/pinctrl/devicetree.c +0
-1@@ -129,9 +129,8 @@ static int dt_to_map_one_config(struct pinctrl *p,if (!np_pctldev ||of_node_is_root (np_pctldev)) {of_node_put (np_pctldev); ret =driver_deferred_probe_check_state (p->dev); -/* keep deferring if modules are enabled unless we've timed out */
drivers/pinctrl/devicetree.c +2
-2-if (IS_ENABLED (CONFIG_MODULES ) && !allow_default && - (ret == -ENODEV )) +/* keep deferring if modules are enabled */ + if (IS_ENABLED(CONFIG_MODULES) &&!allow_default && ret < 0)
ret = -EPROBE_DEFER; return ret; }
Quoted 8 lines
Quoted 3 lines
(you could probably argue calling
driver_deferred_probe_check_statechecking ret at all is silly here, sinceEPROBE_DEFERis the only option you want)Just by looking at it I would've said, yes, this looks like it would fix it. However, upon giving this a try, I see the same pinmux issues as I was seeing before. I'm going to have to dig a little deeper to see if I can find out why that is, because it isn't obvious to me right away.
Oh, nevermind. I managed to somehow mess up the device tree. With that fixed the above also restores the pinmux attachment for the I2C controller and suspend/resume works fine.
Thierry
PATCH v3 2/3 driver core: Use dev_warn() instead of dev_WARN() for deferred_probe_timeout warnings
In reply to John StultzIn commit c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic") and following changes the logic was changes slightly so that if there is no driver to match whats found in the dtb, we wait the sepcified seconds for modules to be loaded by userland, and then timeout, where as previously we'd print "ignoring dependency for device, assuming no driver" and immediately return -ENODEV after initcall_done.
However, in the timeout case (which previously existed but was practicaly un-used without a boot argument), the timeout message uses dev_WARN(). This means folks are now seeing a big backtrace in their boot logs if there a entry in their dts that doesn't have a driver.
To fix this, lets use dev_warn(), instead of dev_WARN() to match the previous error path.
--- drivers/base/dd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/base/dd.c +1 -1
@@ -258,7 +258,7 @@ int driver_deferred_probe_check_state(struct device *dev) }if (!driver_deferred_probe_timeout && initcalls_done) { -dev_WARN (dev,"deferred probe timeout, ignoring dependency" ); +dev_warn (dev,"deferred probe timeout, ignoring dependency" );return -ETIMEDOUT ; }
Signature · 1 line
2.17.1
RE: [PATCH v3 2/3] driver core: Use dev_warn() instead of dev_WARN() for deferred_probe_timeout warnings
Hi John,
Quoted 42 lines
From: John Stultz, Sent: Thursday, April 23, 2020 5:33 AM
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic") and following changes the logic was changes slightly so that if there is no driver to match whats found in the dtb, we wait the sepcified seconds for modules to be loaded by userland, and then timeout, where as previously we'd print "ignoring dependency for device, assuming no driver" and immediately return-ENODEVafterinitcall_done.However, in the timeout case (which previously existed but was practicaly un-used without a boot argument), the timeout message uses
dev_WARN(). This means folks are now seeing a big backtrace in their boot logs if there a entry in their dts that doesn't have a driver.To fix this, lets use
dev_warn(), instead ofdev_WARN()to match the previous error path.Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>
Quoted 1 line
---
Thank you for the patch!
Best regards,
Yoshihiro Shimoda
PATCH v3 3/3 driver core: Ensure wait_for_device_probe() waits until the deferred_probe_timeout fires
In reply to John StultzIn commit c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic"), we set the default driver_deferred_probe_timeout value to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland after initcalls_done is set.
However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the ip_auto_config() had already failed, resulting in NFS to fail. This was due to ip_auto_config() calling wait_for_device_probe() which doesn't wait for the driver_deferred_probe_timeout to fire.
This patch tries to fix the issue by creating a waitqueue for the driver_deferred_probe_timeout, and calling wait_event() to make sure driver_deferred_probe_timeout is zero in wait_for_device_probe() to make sure all the probing is finished.
The downside to this solution is that kernel functionality that uses wait_for_device_probe(), will block until the driver_deferred_probe_timeout fires, regardless of if there is any missing dependencies.
However, the previous patch reverts the default timeout value to zero, so this side-effect will only affect users who specify a driver_deferred_probe_timeout= value as a boot argument, where the additional delay would be beneficial to allow modules to load later during boot.
Thanks to Geert for chasing down that ip_auto_config was why NFS was failing in this case!
--- drivers/base/dd.c | 5 +++++ 1 file changed, 5 insertions(+)
drivers/base/dd.c +5 -0
@@ -226,6 +226,7 @@ DEFINE_SHOW_ATTRIBUTE(deferred_devs);int driver_deferred_probe_timeout;EXPORT_SYMBOL_GPL (driver_deferred_probe_timeout); +static DECLARE_WAIT_QUEUE_HEAD (probe_timeout_waitqueue);static int __init deferred_probe_timeout_setup (char *str) {
@@ -275,6 +276,7 @@ static void deferred_probe_timeout_work_func(struct work_struct *work)list_for_each_entry_safe (private, p, &deferred_probe_pending_list, deferred_probe)dev_info (private->device,"deferred probe pending" ); +wake_up (&probe_timeout_waitqueue); } static DECLARE_DELAYED_WORK(deferred_probe_timeout_work, deferred_probe_timeout_work_func);
@@ -649,6 +651,9 @@ int driver_probe_done(void)*/ void wait_for_device_probe (void ) { +/* wait for probe timeout */ +wait_event (probe_timeout_waitqueue, !driver_deferred_probe_timeout); +/* wait for the deferred probe workqueue to finish */ flush_work (&deferred_probe_work);
Signature · 1 line
2.17.1
Re: [PATCH v3 3/3] driver core: Ensure wait_for_device_probe() waits until the deferred_probe_timeout fires
Quoted 54 lines · John Stultz
In commit
c8c43cee29f6("driver core: Fixdriver_deferred_probe_check_state()logic"), we set the defaultdriver_deferred_probe_timeoutvalue to 30 seconds to allow for drivers that are missing dependencies to have some time so that the dependency may be loaded from userland afterinitcalls_doneis set.However, Yoshihiro Shimoda reported that on his device that expects to have unmet dependencies (due to "optional links" in its devicetree), was failing to mount the NFS root.
In digging further, it seemed the problem was that while the device properly probes after waiting 30 seconds for any missing modules to load, the
ip_auto_config()had already failed, resulting in NFS to fail. This was due toip_auto_config()callingwait_for_device_probe()which doesn't wait for thedriver_deferred_probe_timeoutto fire.This patch tries to fix the issue by creating a waitqueue for the
driver_deferred_probe_timeout, and callingwait_event()to make suredriver_deferred_probe_timeoutis zero inwait_for_device_probe()to make sure all the probing is finished.The downside to this solution is that kernel functionality that uses
wait_for_device_probe(), will block until thedriver_deferred_probe_timeoutfires, regardless of if there is any missing dependencies.However, the previous patch reverts the default timeout value to zero, so this side-effect will only affect users who specify a driver_deferred_probe_timeout= value as a boot argument, where the additional delay would be beneficial to allow modules to load later during boot.
Thanks to Geert for chasing down that
ip_auto_configwas why NFS was failing in this case!Cc "David S. Miller" <[email protected]>Cc Alexey Kuznetsov <[email protected]>Cc Hideaki YOSHIFUJI <[email protected]>Cc Jakub Kicinski <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Rafael J. Wysocki <[email protected]>Cc Rob Herring <[email protected]>Cc Geert Uytterhoeven <[email protected]>Cc Yoshihiro Shimoda <[email protected]>Cc Robin Murphy <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Sudeep Holla <[email protected]>Cc Andy Shevchenko <[email protected]>Cc Naresh Kamboju <[email protected]>Cc Basil Eljuse <[email protected]>Cc Ferry Toth <[email protected]>
Quoted 8 lines
Cc Arnd Bergmann <[email protected]>Cc Anders Roxell <[email protected]>Cc netdev <[email protected]>Reported-by Yoshihiro Shimoda <[email protected]>Tested-by Yoshihiro Shimoda <[email protected]>Fixes c8c43cee29f6 ("driver core: Fix driver_deferred_probe_check_state() logic")Signed-off-by John Stultz <[email protected]>
Let's add my
for v2.
Gr{oetje,eeting}s,
Geert
Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected] In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds