Disregarding the weird global state hiding in this cros_ec_lpc_mec_*()
stuff, it belongs in device probe/remove. We shouldn't assume we can
access hardware resources when the device isn't attached to the driver.
Signed-off-by: Brian Norris <[email protected]>
---
drivers/platform/chrome/cros_ec_lpc.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/platform/chrome/cros_ec_lpc.c b/drivers/platform/chrome/cros_ec_lpc.c
index 7677ab3c0ead..0b6c7c912ec7 100644
--- a/drivers/platform/chrome/cros_ec_lpc.c
+++ b/drivers/platform/chrome/cros_ec_lpc.c
@@ -354,6 +354,9 @@ static int cros_ec_lpc_probe(struct platform_device *pdev)
return -EBUSY;
}
+ cros_ec_lpc_mec_init(EC_HOST_CMD_REGION0,
+ EC_LPC_ADDR_MEMMAP + EC_MEMMAP_SIZE);
+
/*
* Read the mapped ID twice, the first one is assuming the
* EC is a Microchip Embedded Controller (MEC) variant, if the
@@ -456,6 +459,8 @@ static int cros_ec_lpc_remove(struct platform_device *pdev)
cros_ec_unregister(ec_dev);
+ cros_ec_lpc_mec_destroy();
+
return 0;
}
@@ -586,9 +591,6 @@ static int __init cros_ec_lpc_init(void)
return -ENODEV;
}
- cros_ec_lpc_mec_init(EC_HOST_CMD_REGION0,
- EC_LPC_ADDR_MEMMAP + EC_MEMMAP_SIZE);
-
/* Register the driver */
ret = platform_driver_register(&cros_ec_lpc_driver);
if (ret) {
@@ -615,7 +617,6 @@ static void __exit cros_ec_lpc_exit(void)
if (!cros_ec_lpc_acpi_device_found)
platform_device_unregister(&cros_ec_lpc_device);
platform_driver_unregister(&cros_ec_lpc_driver);
- cros_ec_lpc_mec_destroy();
}
module_init(cros_ec_lpc_init);
--
2.38.1.273.g43a17bfeac-goog
This driver often takes on the order of 10ms to start, but in some cases
as much as 600ms [1]. It shouldn't have many cross-device dependencies
to race with, nor racy access to shared state with other drivers, so
this should be a relatively low risk change.
This driver was pinpointed as part of a survey of top slowest initcalls
(i.e., are built in, and probing synchronously) on a lab of ChromeOS
systems.
[1] 600ms was especially surprising to me, so I checked a little deeper.
This driver is used to interface with Embedded Controllers besides just
the traditional laptop power-state controller -- it also interfaces with
some fingerprint readers, which may start up in parallel with the
kernel, or which may not even be present on some SKUs, despite having a
node for it. Thus, our time is wasted just timing out talking to it. At
least we can do that without blocking everyone else.
Signed-off-by: Brian Norris <[email protected]>
---
drivers/platform/chrome/cros_ec_spi.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/platform/chrome/cros_ec_spi.c b/drivers/platform/chrome/cros_ec_spi.c
index 7360b3ff6e4f..21143dba8970 100644
--- a/drivers/platform/chrome/cros_ec_spi.c
+++ b/drivers/platform/chrome/cros_ec_spi.c
@@ -834,6 +834,7 @@ static struct spi_driver cros_ec_driver_spi = {
.name = "cros-ec-spi",
.of_match_table = cros_ec_spi_of_match,
.pm = &cros_ec_spi_pm_ops,
+ .probe_type = PROBE_PREFER_ASYNCHRONOUS,
},
.probe = cros_ec_spi_probe,
.remove = cros_ec_spi_remove,
--
2.38.1.273.g43a17bfeac-goog
This takes on the order of 60ms to probe on some systems, so let it
probe asynchronously. It shouldn't have any dependencies that aren't
handled cleanly.
Signed-off-by: Brian Norris <[email protected]>
---
drivers/platform/chrome/cros_ec_lpc.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/platform/chrome/cros_ec_lpc.c b/drivers/platform/chrome/cros_ec_lpc.c
index 0b6c7c912ec7..527092ffdecb 100644
--- a/drivers/platform/chrome/cros_ec_lpc.c
+++ b/drivers/platform/chrome/cros_ec_lpc.c
@@ -559,6 +559,7 @@ static struct platform_driver cros_ec_lpc_driver = {
.name = DRV_NAME,
.acpi_match_table = cros_ec_lpc_acpi_device_ids,
.pm = &cros_ec_lpc_pm_ops,
+ .probe_type = PROBE_PREFER_ASYNCHRONOUS,
},
.probe = cros_ec_lpc_probe,
.remove = cros_ec_lpc_remove,
--
2.38.1.273.g43a17bfeac-goog
This driver takes on the order of 40ms to start on some systems. It
shouldn't have many cross-device dependencies to race with, nor racy
access to shared state with other drivers, so this should be a
relatively low risk change.
This driver was pinpointed as part of a survey of top slowest initcalls
(i.e., are built in, and probing synchronously) on a lab of ChromeOS
systems.
Signed-off-by: Brian Norris <[email protected]>
---
drivers/platform/chrome/cros_ec_debugfs.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/platform/chrome/cros_ec_debugfs.c b/drivers/platform/chrome/cros_ec_debugfs.c
index 4e63adf083ea..21d973fc6be2 100644
--- a/drivers/platform/chrome/cros_ec_debugfs.c
+++ b/drivers/platform/chrome/cros_ec_debugfs.c
@@ -521,6 +521,7 @@ static struct platform_driver cros_ec_debugfs_driver = {
.driver = {
.name = DRV_NAME,
.pm = &cros_ec_debugfs_pm_ops,
+ .probe_type = PROBE_PREFER_ASYNCHRONOUS,
},
.probe = cros_ec_debugfs_probe,
.remove = cros_ec_debugfs_remove,
--
2.38.1.273.g43a17bfeac-goog
This driver takes on the order of 15ms to start on some systems. Even on
systems where there is no lightbar support, it can take a few
milliseconds just to probe the EC for support. It shouldn't have many
cross-device dependencies to race with, nor racy access to shared state
with other drivers, so this should be a relatively low risk change.
This driver was pinpointed as part of a survey of top slowest initcalls
(i.e., are built in, and probing synchronously) on a lab of ChromeOS
systems.
Signed-off-by: Brian Norris <[email protected]>
---
drivers/platform/chrome/cros_ec_lightbar.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/platform/chrome/cros_ec_lightbar.c b/drivers/platform/chrome/cros_ec_lightbar.c
index 469dfc7a4a03..ff4d619cf924 100644
--- a/drivers/platform/chrome/cros_ec_lightbar.c
+++ b/drivers/platform/chrome/cros_ec_lightbar.c
@@ -601,6 +601,7 @@ static struct platform_driver cros_ec_lightbar_driver = {
.driver = {
.name = DRV_NAME,
.pm = &cros_ec_lightbar_pm_ops,
+ .probe_type = PROBE_PREFER_ASYNCHRONOUS,
},
.probe = cros_ec_lightbar_probe,
.remove = cros_ec_lightbar_remove,
--
2.38.1.273.g43a17bfeac-goog
On Mon, Oct 31, 2022 at 12:55:57PM +0800, Tzung-Bi Shih wrote:
> On Fri, Oct 28, 2022 at 02:14:45PM -0700, Brian Norris wrote:
> > Disregarding the weird global state hiding in this cros_ec_lpc_mec_*()
> > stuff, it belongs in device probe/remove. We shouldn't assume we can
> > access hardware resources when the device isn't attached to the driver.
>
> It's also weird that cros_ec_lpc_mec_destroy() destroies a statically
> allocated mutex[1]. How about let's remove it?
If it makes sense:
https://patchwork.kernel.org/project/chrome-platform/patch/[email protected]/
On Fri, Oct 28, 2022 at 02:14:45PM -0700, Brian Norris wrote:
> Disregarding the weird global state hiding in this cros_ec_lpc_mec_*()
> stuff, it belongs in device probe/remove. We shouldn't assume we can
> access hardware resources when the device isn't attached to the driver.
It's also weird that cros_ec_lpc_mec_destroy() destroies a statically
allocated mutex[1]. How about let's remove it?
[1]: https://elixir.bootlin.com/linux/v6.0/source/drivers/platform/chrome/cros_ec_lpc_mec.c#L152
> @@ -586,9 +591,6 @@ static int __init cros_ec_lpc_init(void)
> return -ENODEV;
> }
>
> - cros_ec_lpc_mec_init(EC_HOST_CMD_REGION0,
> - EC_LPC_ADDR_MEMMAP + EC_MEMMAP_SIZE);
> -
> /* Register the driver */
> ret = platform_driver_register(&cros_ec_lpc_driver);
> if (ret) {
There are 2 more cros_ec_lpc_mec_destroy()s need to be removed [2][3] though.
[2]: https://elixir.bootlin.com/linux/v6.0/source/drivers/platform/chrome/cros_ec_lpc.c#L596
[3]: https://elixir.bootlin.com/linux/v6.0/source/drivers/platform/chrome/cros_ec_lpc.c#L606
On Mon, Oct 31, 2022 at 12:55:57PM +0800, Tzung-Bi Shih wrote:
> There are 2 more cros_ec_lpc_mec_destroy()s need to be removed [2][3] though.
>
> [2]: https://elixir.bootlin.com/linux/v6.0/source/drivers/platform/chrome/cros_ec_lpc.c#L596
> [3]: https://elixir.bootlin.com/linux/v6.0/source/drivers/platform/chrome/cros_ec_lpc.c#L606
Your links / line numbers don't make sense, but I see your point. That
was a bad oversight on my part, sorry.
I see you're probably dropping the destroy() function, so I guess I'll
send v2 without this problem.
Brian
Hello:
This series was applied to chrome-platform/linux.git (for-kernelci)
by Tzung-Bi Shih <[email protected]>:
On Fri, 28 Oct 2022 14:14:45 -0700 you wrote:
> Disregarding the weird global state hiding in this cros_ec_lpc_mec_*()
> stuff, it belongs in device probe/remove. We shouldn't assume we can
> access hardware resources when the device isn't attached to the driver.
>
> Signed-off-by: Brian Norris <[email protected]>
> ---
>
> [...]
Here is the summary with links:
- [1/5] platform/chrome: cros_ec_lpc: Move mec_init/destroy to device probe/remove
(no matching commit)
- [2/5] platform/chrome: cros_ec_lpc: Mark PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/bd88b965ae8c
- [3/5] platform/chrome: cros_ec_debugfs: Set PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/692a68ad7f3c
- [4/5] platform/chrome: cros_ec_lightbar: Set PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/873ab3e886b5
- [5/5] platform/chrome: cros_ec_spi: Set PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/015e4b05c377
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
Hello:
This series was applied to chrome-platform/linux.git (for-next)
by Tzung-Bi Shih <[email protected]>:
On Fri, 28 Oct 2022 14:14:45 -0700 you wrote:
> Disregarding the weird global state hiding in this cros_ec_lpc_mec_*()
> stuff, it belongs in device probe/remove. We shouldn't assume we can
> access hardware resources when the device isn't attached to the driver.
>
> Signed-off-by: Brian Norris <[email protected]>
> ---
>
> [...]
Here is the summary with links:
- [1/5] platform/chrome: cros_ec_lpc: Move mec_init/destroy to device probe/remove
(no matching commit)
- [2/5] platform/chrome: cros_ec_lpc: Mark PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/bd88b965ae8c
- [3/5] platform/chrome: cros_ec_debugfs: Set PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/692a68ad7f3c
- [4/5] platform/chrome: cros_ec_lightbar: Set PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/873ab3e886b5
- [5/5] platform/chrome: cros_ec_spi: Set PROBE_PREFER_ASYNCHRONOUS
https://git.kernel.org/chrome-platform/c/015e4b05c377
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html