coccinelle: api/devm_platform_ioremap_resource: remove useless script
Subsystems (2)
PATCH 1/1 coccinelle: api/devm_platform_ioremap_resource: remove useless script
While it is useful for new drivers to use devm_platform_ioremap_resource, this script is currently used to spam maintainers, often updating very old drivers. The net benefit is the removal of 2 lines of code in the driver but the review load for the maintainers is huge. As of now, more that 560 patches have been sent, some of them obviously broken, as in:
https://lore.kernel.org/lkml/[email protected]/
Remove the script to reduce the spam.
--- .../api/devm_platform_ioremap_resource.cocci | 60 ------------------- 1 file changed, 60 deletions(-) delete mode 100644 scripts/coccinelle/api/devm_platform_ioremap_resource.cocci
scripts/coccinelle/api/devm_platform_ioremap_resource.cocci deleted +0 -60
@@ -1,60 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -/// Use devm_platform_ioremap_resource helper which wraps -/// platform_get_resource() and devm_ioremap_resource() together. -/// -// Confidence: High -// Copyright: (C) 2019 Himanshu Jha GPLv2. -// Copyright: (C) 2019 Julia Lawall, Inria/LIP6. GPLv2. -// Keywords: platform_get_resource, devm_ioremap_resource, -// Keywords: devm_platform_ioremap_resource - -virtual patch -virtual report - -@r depends on patch && !report@ -expression e1, e2, arg1, arg2, arg3; -identifier id; -@@ - -( -- id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -| -- struct resource *id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -) - ... when != id -- e1 = devm_ioremap_resource(arg3, id); -+ e1 = devm_platform_ioremap_resource(arg1, arg2); - ... when != id -? id = e2 - -@r1 depends on patch && !report@ -identifier r.id; -type T; -@@ - -- T *id; - ...when != id - -@r2 depends on report && !patch@ -identifier id; -expression e1, e2, arg1, arg2, arg3; -position j0; -@@ - -( - id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -| - struct resource *id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -) - ... when != id - e1@j0 = devm_ioremap_resource(arg3, id); - ... when != id -? id = e2 - -@script:python depends on report && !patch@ -e1 << r2.e1; -j0 << r2.j0; -@@ - -msg = "WARNING: Use devm_platform_ioremap_resource for %s" % (e1) -coccilib.report.print_report(j0[0], msg)
Signature · 1 line
2.21.0
Quoted 9 lines · Alexandre Belloni
While it is useful for new drivers to use
devm_platform_ioremap_resource, this script is currently used to spam maintainers, often updating very old drivers. The net benefit is the removal of 2 lines of code in the driver but the review load for the maintainers is huge. As of now, more that 560 patches have been sent, some of them obviously broken, as in:https://lore.kernel.org/lkml/[email protected]/
Remove the script to reduce the spam.
OK.
Quoted 77 lines
Signed-off-by Alexandre Belloni <[email protected]>--- .../api/devm_platform_ioremap_resource.cocci | 60 ------------------- 1 file changed, 60 deletions(-) delete mode 100644 scripts/coccinelle/api/devm_platform_ioremap_resource.cocciscripts/coccinelle/api/devm_platform_ioremap_resource.cocci deleted +0
-61@@ -1,60 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -/// Use devm_platform_ioremap_resource helper which wraps -/// platform_get_resource() and devm_ioremap_resource() together. -/// -// Confidence: High -// Copyright: (C) 2019 Himanshu Jha GPLv2. -// Copyright: (C) 2019 Julia Lawall, Inria/LIP6. GPLv2. -// Keywords: platform_get_resource, devm_ioremap_resource, -// Keywords: devm_platform_ioremap_resource - -virtual patch -virtual report - -@r depends on patch && !report@ -expression e1, e2, arg1, arg2, arg3; -identifier id; -@@ - -( -- id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -| -- struct resource *id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -) - ... when != id -- e1 = devm_ioremap_resource(arg3, id); -+ e1 = devm_platform_ioremap_resource(arg1, arg2); - ... when != id -? id = e2 - -@r1 depends on patch && !report@ -identifier r.id; -type T; -@@ - -- T *id; - ...when != id - -@r2 depends on report && !patch@ -identifier id; -expression e1, e2, arg1, arg2, arg3; -position j0; -@@ - -( - id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -| - struct resource *id = platform_get_resource(arg1, IORESOURCE_MEM, arg2); -) - ... when != id - e1@j0 = devm_ioremap_resource(arg3, id); - ... when != id -? id = e2 - -@script:python depends on report && !patch@ -e1 << r2.e1; -j0 << r2.j0; -@@ - -msg = "WARNING: Use devm_platform_ioremap_resource for %s" % (e1) -coccilib.report.print_report(j0[0], msg) --2.21.0
Quoted 11 lines · Alexandre Belloni
While it is useful for new drivers to use
devm_platform_ioremap_resource,
this script is currently used to spam maintainers, often updating very old
drivers. The net benefit is the removal of 2 lines of code in the driver
but the review load for the maintainers is huge. As of now, more that 560
patches have been sent, some of them obviously broken, as in:
Quoted 5 lines
https://lore.kernel.org/lkml/[email protected]/
Remove the script to reduce the spam.
Signed-off-by Alexandre Belloni <[email protected]>
I think part of the issue is that the script reports a WARNING for something
that is definitely correct code, and could instead be simply toned down.
Anyway, FWIW:
M.
Signature · 1 line
Jazz is not dead. It just smells funny...
Quoted 1 line
I think part of the issue is that the script reports a WARNING
How much does this information influence really the stress tolerance and change resistance (or acceptance) for the presented collateral evolution?
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/scripts/coccinelle/api/devm_platform_ioremap_resource.cocci
Quoted 1 line
for something that is definitely correct code,
Can related software improvement possibilities be taken into account again under other circumstances?
Quoted 1 line
and could instead be simply toned down.
Does this view mean that the mentioned script for the semantic patch language should get another chance for integration?
Quoted 3 lines
Anyway, FWIW:
Acked-by Marc Zyngier <[email protected]>
Would you like to share any more constructive feedback?
Will similar source file mass updates be better picked up by other well-known Linux developers?
Regards,
Markus
Quoted 6 lines · Markus Elfring
Quoted 2 lines
I think part of the issue is that the script reports a WARNING
How much does this information influence really the stress tolerance and change resistance (or acceptance) for the presented collateral evolution?
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/scripts/coccinelle/api/devm_platform_ioremap_resource.cocci
-ENOPARSE.
Quoted 4 lines
Quoted 1 line
for something that is definitely correct code,
Can related software improvement possibilities be taken into account again under other circumstances?
These patches provide no improvement whatsoever. As pointed out, they mostly introduce bugs.
Quoted 4 lines
Quoted 1 line
and could instead be simply toned down.
Does this view mean that the mentioned script for the semantic patch language should get another chance for integration?
Providing Coccinelle scripts that scream about perfectly valid code is pointless, and the result is actively harmful.
If said script was providing a correct semantic patch instead of being an incentive for people to churn untested patches that span the whole tree, that'd be a different story. But that's not what this is about.
Quoted 5 lines
Quoted 3 lines
Anyway, FWIW:
Acked-by Marc Zyngier <[email protected]>Would you like to share any more constructive feedback?
No.
Quoted 2 lines
Will similar source file mass updates be better picked up by other well-known Linux developers?
Certainly not for the subsystems I maintain.
M.
Signature · 1 line
Jazz is not dead, it just smells funny.
Quoted 2 lines · Marc Zyngier
Providing Coccinelle scripts that scream about perfectly valid code is pointless, and the result is actively harmful.
Doubtful.
If the new code is smaller object code and correct than the conversion is worthwhile.
fyi:
There are already ~450 uses of this function and maybe ~800 possible additional conversions.
Quoted 3 lines
If said script was providing a correct semantic patch instead of being an incentive for people to churn untested patches that span the whole tree, that'd be a different story.
Right.
Quoted 9 lines · Joe Perches
Quoted 2 lines · Marc Zyngier
Providing Coccinelle scripts that scream about perfectly valid code is pointless, and the result is actively harmful.
Doubtful.
If the new code is smaller object code and correct than the conversion is worthwhile.
I agree.
We use multi-platform defconfig.
I always appreciate the code refactoring
that reduces the object size.
Quoted 12 lines
fyi:
There are already ~450 uses of this function and maybe ~800 possible additional conversions.
Quoted 3 lines
If said script was providing a correct semantic patch instead of being an incentive for people to churn untested patches that span the whole tree, that'd be a different story.
Right.
Alexandre Belloni used
https://lore.kernel.org/lkml/[email protected]/
as a reference, but this is not the output from coccicheck. The patch author just created a wrong patch by hand.
The deleted semantic patch supports MODE=patch,
which creates a correct patch, and is useful.
Signature · 2 lines
Best Regards Masahiro Yamada
Quoted 2 lines
I always appreciate the code refactoring
that reduces the object size.
Would you like to compare effects around conversions for the mentioned wrapper function any more?
Regards,
Markus
Quoted 7 lines · Masahiro Yamada
Quoted 1 line · Joe Perches
On Sat, 2019-10-19 at 21:43 +0100, Marc Zyngier wrote:
Alexandre Belloni used
https://lore.kernel.org/lkml/[email protected]/
as a reference, but this is not the output from coccicheck. The patch author just created a wrong patch by hand.
Exactly. Removal of the script is a mistake. Like I said before is a healing (incorrect by the way!) by symptoms.
Quoted 2 lines
The deleted semantic patch supports MODE=patch,
which creates a correct patch, and is useful.
Right!
Signature · 2 lines
With Best Regards, Andy Shevchenko
Quoted 9 lines · Andy Shevchenko
Quoted 7 lines · Masahiro Yamada
Quoted 1 line · Joe Perches
On Sat, 2019-10-19 at 21:43 +0100, Marc Zyngier wrote:
Alexandre Belloni used
https://lore.kernel.org/lkml/[email protected]/
as a reference, but this is not the output from coccicheck. The patch author just created a wrong patch by hand.
Quoted 7 lines
Exactly. Removal of the script is a mistake. Like I said before is a healing (incorrect by the way!) by symptoms.
Quoted 2 lines
The deleted semantic patch supports MODE=patch,
which creates a correct patch, and is useful.Right!
I ran it on the version of Linux that still has the script:
fe7d2c23d748e4206f4bef9330d0dff9abed7411
and managed to compile 341 of the generated files in the time I had available, and all compiled successfully. I can let it run again, and see how it goes for the rest. Perhaps it would be acceptable if there was no report, and people would be forced to use the generated patch?
If someone is writing lots of patches on this issue by hand, then perhaps they don't have make coccicheck to produce patches, and then would overlook this case completely.
If it would be helpful, I could group the generated patches by maintainer or by subdirectory and send them out, if it would be easier to review them all at once.
Anyway, the rule is not in the kernel at the moment. For it's future, I'm open to whatever people find best. Personally, I prefer when same things are done in the same way - it makes the code easier to understand and makes it simpler to address other issues when they arise.
julia
Hi Julia
Quoted 20 lines · Julia Lawall
Quoted 15 lines · Andy Shevchenko
Quoted 7 lines · Masahiro Yamada
Quoted 1 line · Joe Perches
On Sat, 2019-10-19 at 21:43 +0100, Marc Zyngier wrote:
Alexandre Belloni used
https://lore.kernel.org/lkml/[email protected]/
as a reference, but this is not the output from coccicheck. The patch author just created a wrong patch by hand.Exactly. Removal of the script is a mistake. Like I said before is a healing (incorrect by the way!) by symptoms.
Quoted 2 lines
The deleted semantic patch supports MODE=patch,
which creates a correct patch, and is useful.
Quoted 8 lines
Quoted 1 line
Right!
I ran it on the version of Linux that still has the script:
fe7d2c23d748e4206f4bef9330d0dff9abed7411and managed to compile 341 of the generated files in the time I had available, and all compiled successfully.
Yeah, this semantic patch did the correct conversion as its header part showed the confidence.
// Confidence: High
Quoted 3 lines
I can let it run again, and see
how it goes for the rest. Perhaps it would be acceptable if there was no report, and people would be forced to use the generated patch?
I do not think this is the right thing.
MODE=report is the default, and it is fine.
Quoted 8 lines
If someone is writing lots of patches on this issue by hand, then perhaps they don't have make coccicheck to produce patches, and then would overlook this case completely.
If it would be helpful, I could group the generated patches by maintainer or by subdirectory and send them out, if it would be easier to review them all at once.
Yes, please.
Subsystem maintainers trust you,
so I think it will make things move smoothly.
After converting most of files,
I want 283ea345934d277e30c841c577e0e2142b4bfcae reverted.
Quoted 5 lines
Anyway, the rule is not in the kernel at the moment. For it's future, I'm open to whatever people find best. Personally, I prefer when same things are done in the same way - it makes the code easier to understand and makes it simpler to address other issues when they arise.
We always did the same things in the same way except commit 283ea345934d277e30c841c577e0e2142b4bfcae
Signature · 2 lines
Best Regards Masahiro Yamada
Quoted 55 lines · Masahiro Yamada
Hi Julia
Quoted 20 lines · Julia Lawall
Quoted 15 lines · Andy Shevchenko
Quoted 7 lines · Masahiro Yamada
Quoted 1 line · Joe Perches
On Sat, 2019-10-19 at 21:43 +0100, Marc Zyngier wrote:
Alexandre Belloni used
https://lore.kernel.org/lkml/[email protected]/
as a reference, but this is not the output from coccicheck. The patch author just created a wrong patch by hand.Exactly. Removal of the script is a mistake. Like I said before is a healing (incorrect by the way!) by symptoms.
Quoted 2 lines
The deleted semantic patch supports MODE=patch,
which creates a correct patch, and is useful.Quoted 8 lines
Quoted 1 line
Right!
I ran it on the version of Linux that still has the script:
fe7d2c23d748e4206f4bef9330d0dff9abed7411and managed to compile 341 of the generated files in the time I had available, and all compiled successfully.
Yeah, this semantic patch did the correct conversion as its header part showed the confidence.
// Confidence: High
Quoted 3 lines
I can let it run again, and see
how it goes for the rest. Perhaps it would be acceptable if there was no report, and people would be forced to use the generated patch?I do not think this is the right thing.
MODE=report is the default, and it is fine.Quoted 8 lines
If someone is writing lots of patches on this issue by hand, then perhaps they don't have make coccicheck to produce patches, and then would overlook this case completely.
If it would be helpful, I could group the generated patches by maintainer or by subdirectory and send them out, if it would be easier to review them all at once.
Quoted 7 lines
Yes, please.
Subsystem maintainers trust you,
so I think it will make things move smoothly.After converting most of files,
I want283ea345934d277e30c841c577e0e2142b4bfcaereverted.
OK. I got 477 of the files to compile directly. I can send patches on them, and then look into the issues on the remaining ones (probably configuration issues).
julia
Quoted 19 lines
Quoted 7 lines
Anyway, the rule is not in the kernel at the moment. For it's future, I'm open to whatever people find best. Personally, I prefer when same things are done in the same way - it makes the code easier to understand and makes it simpler to address other issues when they arise.
We always did the same things in the same way except commit
283ea345934d277e30c841c577e0e2142b4bfcae--
Best Regards
Masahiro Yamada
Quoted 3 lines
If said script was providing a correct semantic patch instead of being an incentive for people to churn untested patches that span the whole tree, that'd be a different story. But that's not what this is about.
What is the actual incorrectness with the script?
An option could be to adjust the rule such that it can be run with an extra command line option, like -D developer but is not run by default by make coccicheck.
thanks,
julia
Quoted 6 lines · Julia Lawall
Quoted 4 lines
If said script was providing a correct semantic patch instead of being an incentive for people to churn untested patches that span the whole tree, that'd be a different story. But that's not what this is about.
What is the actual incorrectness with the script?
The first thing is that it spits out a "WARNING", which is almost universally interpreted as something that needs addressing. In this case, it really doesn't. The suggested helper is only icing sugar, and the original code is perfectly fine.
The second thing is that it results in people posting patches they don't even compile, let alone test. There would be a good chance for these patches to be correct if the script was directly generating them, but that's unfortunately not the case.
Quoted 3 lines
An option could be to adjust the rule such that it can be run with an extra command line option, like -D developer but is not run by default by make coccicheck.
Maybe. I'm not sure this will deter people from running these scripts and sending broken patches anyway. No matter how many safeguards you put, people will still post broken patches just because they can.
M.
Signature · 1 line
Jazz is not dead, it just smells funny.
Quoted 1 line
Quoted 1 line
Quoted 1 line
I think part of the issue is that the script reports a WARNING
Would anybody like to change this category to “INFO”?
Quoted 5 lines
Quoted 3 lines
How much does this information influence really the stress tolerance and change resistance (or acceptance) for the presented collateral evolution?
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/scripts/coccinelle/api/devm_platform_ioremap_resource.cocci
-ENOPARSE.
* Automated processes can trigger also big amounts of possible adjustments.
* The software development capacity will vary for affected components during the years.
* Implementing changes is a recurring project management task, isn't it?
Quoted 6 lines
Quoted 4 lines
Quoted 1 line
for something that is definitely correct code,
Can related software improvement possibilities be taken into account again under other circumstances?
These patches provide no improvement whatsoever.
* Do you find information from the description of a corresponding commit 7945f929f1a77a1c8887a97ca07f87626858ff42 ("drivers: provide devm_platform_ioremap_resource()") reasonable?
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/base/platform.c
* How do you think about to compare any differences with software build results?
Quoted 1 line
As pointed out, they mostly introduce bugs.
Would you like to check any error statistics in more detail?
Quoted 1 line
Providing Coccinelle scripts that scream about perfectly valid code is pointless,
They usually point opportunities out for further collateral evolution, don't they?
Quoted 1 line
and the result is actively harmful.
You might not like some changes for a while.
Quoted 1 line
If said script was providing a correct semantic patch
I got the impression that this can also happen often enough. Would you like to check the concrete transformation failure rate once more?
Quoted 2 lines
instead of being an incentive for people to churn untested patches that span the whole tree, that'd be a different story.
Various developers got motivated to achieve something (possible improvements?) also by the means of available software analysis tools.
Mistakes can then happen as usual during such adjustment attempts.
Quoted 1 line
But that's not what this is about.
I guess that your software development concerns can be clarified a bit more.
Regards,
Markus
Quoted 1 line
While it is useful for new drivers to use
devm_platform_ioremap_resource,
This is nice.
Quoted 1 line
this script is currently used to spam maintainers,
This view is unfortunate.
Do we stumble on a target conflict again?
Quoted 1 line
often updating very old drivers.
This can also happen.
Quoted 1 line
The net benefit is the removal of 2 lines of code in the driver
Additional effects can be reconsidered, can't they?
Quoted 1 line
but the review load for the maintainers is huge.
Does collateral evolution trigger a remarkable amount of changes occasionally?
How will such feedback influence the development and integration of further scripts for the semantic patch language (Coccinelle software)?
Regards,
Markus
Quoted 12 lines · Markus Elfring
Quoted 1 line
While it is useful for new drivers to use
devm_platform_ioremap_resource,This is nice.
Quoted 1 line
this script is currently used to spam maintainers,
This view is unfortunate.
Do we stumble on a target conflict again?
Quoted 9 lines
Quoted 1 line
often updating very old drivers.
This can also happen.
Quoted 1 line
The net benefit is the removal of 2 lines of code in the driver
Additional effects can be reconsidered, can't they?
What are the additional effects? What is the end goal of converting all the existing drivers to devm_platform_ioremap_resource? The existing code is currently always correct and it is difficult to see how this would lead to any bug avoidance in the long term.
Quoted 4 lines
Quoted 1 line
but the review load for the maintainers is huge.
Does collateral evolution trigger a remarkable amount of changes occasionally?
This is not an evolution, it is unnecessary churn. Those patches have no benefit and eat up very valuable reviewer time.
Quoted 4 lines
How will such feedback influence the development and integration of further scripts for the semantic patch language (Coccinelle software)?
There are a few other scripts that have no added value when applied to existing code, like ptr_ret.cocci.
Signature · 3 lines
Alexandre Belloni, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
Quoted 1 line
What are the additional effects?
I suggest to take another look at the commit 7945f929f1a77a1c8887a97ca07f87626858ff42 ("drivers: provide devm_platform_ioremap_resource()" from 2019-02-20) which triggered the discussed software evolution.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/base/platform.c
Quoted 1 line
What is the end goal of converting all the existing drivers to
devm_platform_ioremap_resource?
It was accepted by well-known Linux developers to put two function calls into another wrapper function.
Quoted 2 lines
This is not an evolution, it is unnecessary churn. Those patches have no benefit and eat up very valuable reviewer time.
I am curious if other contributors would like to describe more variants of software development opinions in affected areas.
Quoted 5 lines
Quoted 2 lines
How will such feedback influence the development and integration of further scripts for the semantic patch language (Coccinelle software)?
There are a few other scripts that have no added value when applied to existing code, like
ptr_ret.cocci.
Would you like to clarify concerns around such source code transformation approaches in more detail?
Regards,
Markus
sob., 19 paź 2019 o 14:09 Alexandre Belloni
<[email protected]> napisał(a):
Quoted 42 lines · Markus Elfring
Quoted 12 lines
Quoted 1 line
While it is useful for new drivers to use
devm_platform_ioremap_resource,This is nice.
Quoted 1 line
this script is currently used to spam maintainers,
This view is unfortunate.
Do we stumble on a target conflict again?
Quoted 9 lines
Quoted 1 line
often updating very old drivers.
This can also happen.
Quoted 1 line
The net benefit is the removal of 2 lines of code in the driver
Additional effects can be reconsidered, can't they?
What are the additional effects? What is the end goal of converting all the existing drivers to
devm_platform_ioremap_resource? The existing code is currently always correct and it is difficult to see how this would lead to any bug avoidance in the long term.Quoted 4 lines
Quoted 1 line
but the review load for the maintainers is huge.
Does collateral evolution trigger a remarkable amount of changes occasionally?
This is not an evolution, it is unnecessary churn. Those patches have no benefit and eat up very valuable reviewer time.
Quoted 4 lines
How will such feedback influence the development and integration of further scripts for the semantic patch language (Coccinelle software)?
Quoted 7 lines
There are a few other scripts that have no added value when applied to existing code, like
ptr_ret.cocci.--
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
Hi Alexandre,
Markus has been black-listed by several core maintainers already, I think you're wasting your time arguing. WRT the patch: when introducing this wrapper, I definitely didn't expect people to send hundreds of often wrong patches based on coccinelle reports, so I guess removing the script is correct.
Bart
Quoted 1 line
Markus has been black-listed by several core maintainers already,
I am still curious if this communication filter will ever be adjusted in more positive directions.
Quoted 1 line
I think you're wasting your time arguing.
I hope that also this software development discussion can become more constructive.
Quoted 2 lines
WRT the patch: when introducing this wrapper, I definitely didn't expect people to send hundreds of often wrong patches based on coccinelle reports,
The reality can provide various surprises.
Quoted 1 line
so I guess removing the script is correct.
I suggest to reconsider this conclusion once more.
The application of SmPL script variants can be continued despite of the recently committed file removal.
The constraints for the usage of available scripts for the semantic patch language are explained to some degree.
https://bottest.wiki.kernel.org/coccicheck
Regards,
Markus