Skip to contentSkip to conversation

Driver and/or module versions

Ksummit-discuss Leon Romanovsky 8 participants 15 messages Jun 25, 2017 – Jul 3, 2017

lore
Share & tools
lore
#0 lore raw

There is a steady flow of patches which bump driver and module versions.

The differences in views add [1] an email burden, which can be easily avoided by adding policy into CodingStyle guide.

As an example of such patches: "Succeeded" [2] vs. "Failed" [3].

I think that maintainers summit with major code "pushers" is the right venue
to get the answers on the following questions:
* Can we come with unified policy about those patches?
* Can we remove them? If yes, is it possible to replace MODULE_VERSION in all patches to point to kernel version?
* Does such change fail under category "UAPI changes and we can't change" or not?

Thanks

[1] https://patchwork.kernel.org/patch/9735855/
[2] https://patchwork.ozlabs.org/patch/779864/
[3] https://patchwork.kernel.org/patch/7490321/

#1 lore raw
Quoted 3 lines · Leon Romanovsky

There is a steady flow of patches which bump driver and module versions.

* Can we come with unified policy about those patches?

We pretty much *have* a unified policy, it's just that I think rdma is crazy.

The unified policy is pretty much that version codes do not matter, do not exist, and do not get updated.

Things are supposed to be backwards and forwards compatible, because we don't accept breakage in user space anyway. So versioning is pointless, and only causes problems.

It causes problems not just because of the conflict issues, but because it's fundamentally wrong, and makes driver writers think that it's ok to change interfaces and use versioning to show they changed. It's *not* OK.

Sometimes you have feature masks (which just mean that it's ok to _add_ interfaces rather than change them, and make it possible for user space to check if the new interface exists), but even that is generally the exception rather than the rule and should be used very very carefully and preferably not at all.

Linus

#2 lore raw
Quoted 9 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

There is a steady flow of patches which bump driver and module versions.

* Can we come with unified policy about those patches?

We pretty much *have* a unified policy, it's just that I think rdma is crazy.

The unified policy is pretty much that version codes do not matter, do not exist, and do not get updated.

I'm pretty sure that not everyone is aware of it. Those examples of "accepted" patches are not from RDMA and maybe it will be awkward to hear, but in RDMA subsystem we don't allow such patches [1].

My attempt to enforce such policy in other subsystem didn't work [2] and in addition to that failed attempt, I bombarded internally with customers' requests who have cargo cult to update those versions anyway ("like everyone else"). This is why I'm proposing to ban those patches and remove their exposure for the drivers/* completely.

Quoted 9 lines

Things are supposed to be backwards and forwards compatible, because we don't accept breakage in user space anyway. So versioning is pointless, and only causes problems.

It causes problems not just because of the conflict issues, but because it's fundamentally wrong, and makes driver writers think that it's ok to change interfaces and use versioning to show they changed. It's *not* OK.

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

Quoted 6 lines

Sometimes you have feature masks (which just mean that it's ok to _add_ interfaces rather than change them, and make it possible for user space to check if the new interface exists), but even that is generally the exception rather than the rule and should be used very very carefully and preferably not at all.

[1] https://patchwork.kernel.org/patch/9735855/
[2] https://lkml.org/lkml/2016/7/15/17

Quoted 2 lines

Linus

#3 lore raw
Quoted 3 lines · Leon Romanovsky

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

I suspect they just haven't caused problems elsewhere.

And honestly, they aren't that common. Grepping for MODULE_VERSION and DRIVER_VERSION shows more than I would like, but we're talking hundreds, not thousands (and lots of them are basically dead, stale code - ATA etc, or just joke names - usb).

Linus

#4 lore raw
Quoted 11 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

I suspect they just haven't caused problems elsewhere.

And honestly, they aren't that common. Grepping for MODULE_VERSION and DRIVER_VERSION shows more than I would like, but we're talking hundreds, not thousands (and lots of them are basically dead, stale code - ATA etc, or just joke names - usb).

I see it differently, from my point of view, it is pretty active, especially in the subsystems where new drivers are added more often, but if you say that the policy exists, it is enough for me.

Thanks

Quoted 2 lines

Linus

#5 lore raw

Hi,

Quoted 8 lines · Leon Romanovsky
Quoted 7 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

I suspect they just haven't caused problems elsewhere.

Quoted 9 lines
Quoted 4 lines

And honestly, they aren't that common. Grepping for MODULE_VERSION and DRIVER_VERSION shows more than I would like, but we're talking hundreds, not thousands (and lots of them are basically dead, stale code - ATA etc, or just joke names - usb).

I see it differently, from my point of view, it is pretty active, especially in the subsystems where new drivers are added more often, but if you say that the policy exists, it is enough for me.

A while ago, I went and removed all of them from the RTC subsystem.

Signature · 3 lines
Alexandre Belloni, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
#6 lore raw
Quoted 16 lines · Alexandre Belloni

Hi,

Quoted 13 lines · Leon Romanovsky
Quoted 11 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

I suspect they just haven't caused problems elsewhere.

And honestly, they aren't that common. Grepping for MODULE_VERSION and DRIVER_VERSION shows more than I would like, but we're talking hundreds, not thousands (and lots of them are basically dead, stale code - ATA etc, or just joke names - usb).

Quoted 6 lines
Quoted 4 lines

I see it differently, from my point of view, it is pretty active, especially in the subsystems where new drivers are added more often, but if you say that the policy exists, it is enough for me.

A while ago, I went and removed all of them from the RTC subsystem.

I assume that you are referring to the commit fa5691131a87 ("rtc: remove useless DRV_VERSION").

Let's see how similar patch will be accepted in RDMA, netdev communities.

Thanks

Quoted 7 lines

--
Alexandre Belloni, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

#7 lore raw
Quoted 22 lines · Leon Romanovsky
Quoted 21 lines · Alexandre Belloni

Hi,

Quoted 17 lines · Leon Romanovsky
Quoted 11 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

I suspect they just haven't caused problems elsewhere.

And honestly, they aren't that common. Grepping for MODULE_VERSION and DRIVER_VERSION shows more than I would like, but we're talking hundreds, not thousands (and lots of them are basically dead, stale code - ATA etc, or just joke names - usb).

I see it differently, from my point of view, it is pretty active, especially in the subsystems where new drivers are added more often, but if you say that the policy exists, it is enough for me.

Quoted 5 lines
Quoted 1 line

A while ago, I went and removed all of them from the RTC subsystem.

I assume that you are referring to the commit fa5691131a87 ("rtc: remove useless DRV_VERSION").

Let's see how similar patch will be accepted in RDMA, netdev communities.

The only version that matters is CONFIG_LOCALVERSION_AUTO=y.

Gr{oetje,eeting}s,

Geert

--
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

#8 lore raw
In reply to Leon Romanovsky
Quoted 8 lines · Leon Romanovsky
Quoted 7 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

I suspect they just haven't caused problems elsewhere.

Quoted 7 lines
Quoted 4 lines

And honestly, they aren't that common. Grepping for MODULE_VERSION and DRIVER_VERSION shows more than I would like, but we're talking hundreds, not thousands (and lots of them are basically dead, stale code - ATA etc, or just joke names - usb).

I see it differently, from my point of view, it is pretty active, especially in the subsystems where new drivers are added more often,

It takes a few years for companies to realize that the version numbers mean nothing. So give it time, it will slow down. Or you can just delete all of them at once, making it easier for you in the end.

thanks,

greg k-h

#9 lore raw
Quoted 14 lines · Greg KH
Quoted 13 lines · Leon Romanovsky
Quoted 11 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

Great, so why are we continuing to allow patches with MODULE_VERSION and DRIVER_VERSION changes?

I suspect they just haven't caused problems elsewhere.

And honestly, they aren't that common. Grepping for MODULE_VERSION and DRIVER_VERSION shows more than I would like, but we're talking hundreds, not thousands (and lots of them are basically dead, stale code - ATA etc, or just joke names - usb).

Quoted 6 lines
Quoted 2 lines

I see it differently, from my point of view, it is pretty active, especially in the subsystems where new drivers are added more often,

It takes a few years for companies to realize that the version numbers mean nothing. So give it time, it will slow down. Or you can just delete all of them at once, making it easier for you in the end.

OK, let's start in small scale and limit myself to MODULE_VERSION at this stage.
https://patchwork.kernel.org/patch/9808651/

Thanks

Quoted 4 lines

thanks,

greg k-h

#10 lore raw
Quoted 2 lines · Leon Romanovsky

OK, let's start in small scale and limit myself to MODULE_VERSION at this stage.
https://patchwork.kernel.org/patch/9808651/

Isn't it the case that something like this will actually break the ABI?

I assume that a bunch of different companies have scripts that pull out the module version out of the module and make sure that it's at some value that works with the rest of their tools, right?

I also have to assume that these scripts will horribly break when the value just disappears.

Signature · 2 lines
Thanks,
Sasha
#11 lore raw
Quoted 5 lines · Alexander (Sasha Levin)
Quoted 2 lines · Leon Romanovsky

OK, let's start in small scale and limit myself to MODULE_VERSION at this stage.
https://patchwork.kernel.org/patch/9808651/

Isn't it the case that something like this will actually break the ABI?

From the responses, the answer is NO, it is not ABI.

Quoted 4 lines

I assume that a bunch of different companies have scripts that pull out the module version out of the module and make sure that it's at some value that works with the rest of their tools, right?

I also have to assume that these scripts will horribly break when the value just disappears.

It is the same as to rely on console prints as reliable source for those scripts. It is broken by design.

Quoted 5 lines

--

Thanks,
Sasha

#12 lore raw
In reply to Linus Torvalds
Quoted 10 lines · Linus Torvalds
Quoted 3 lines · Leon Romanovsky

There is a steady flow of patches which bump driver and module versions.

* Can we come with unified policy about those patches?

We pretty much *have* a unified policy, it's just that I think rdma is crazy.

The unified policy is pretty much that version codes do not matter, do not exist, and do not get updated.

I make the point internally that without a stable-abi, the driver is not stand-alone, and so the "version" is easily shown to not be independent from the kernel version (or the distro package version).

Quoted 9 lines

Things are supposed to be backwards and forwards compatible, because we don't accept breakage in user space anyway. So versioning is pointless, and only causes problems.

It causes problems not just because of the conflict issues, but because it's fundamentally wrong, and makes driver writers think that it's ok to change interfaces and use versioning to show they changed. It's *not* OK.

Quoted 6 lines

Sometimes you have feature masks (which just mean that it's ok to _add_ interfaces rather than change them, and make it possible for user space to check if the new interface exists), but even that is generally the exception rather than the rule and should be used very very carefully and preferably not at all.

While the module version is clearly a flawed concept for Linux drivers, I do think we should discuss the problem they are intended to address:
knowing what's in a particular instance of a driver.

I see three categories of changes vendors need to be able to track:

1. Bug Fixes
2. Performance enhancements
3. New Features

The first is easily managed through the stable kernel process. Vendors ensure bug fixes are pushed back to stable, distros pick them up from there, done.

The second is split into two categories. Many performance regressions can be accommodated via the stable kernel rules. Regressions with large mitigation patches and performance enhancements currently need to be tracked independently from kernel version as they are backported to distro / product kernels.

New features also fall into the independent tracking bucket, although your point about feature masks could reduce that need. Is there a definitive mechanism for the feature mask approach? I see a lot of sysfs_filename:value key:value pairs for this kind of thing.

Signature · 2 lines
Darren Hart
VMware Open Source Technology Center
#13 lore raw
Quoted 4 lines · Darren Hart

New features also fall into the independent tracking bucket, although your point about feature masks could reduce that need. Is there a definitive mechanism for the feature mask approach? I see a lot of sysfs_filename:value key:value pairs for this kind of thing.

Adding those sysfs attributes seems like exactly the thing that people will keep forgetting to do, as there is no (real) functionality depending on them.

I doubt there is any better 'description' of the 'state' of the driver than SHA of the topmost commit + the tree it's related to.

Signature · 2 lines
Jiri Kosina
SUSE Labs
#14 lore raw
Quoted 7 lines · Jiri Kosina
Quoted 4 lines · Darren Hart

New features also fall into the independent tracking bucket, although your point about feature masks could reduce that need. Is there a definitive mechanism for the feature mask approach? I see a lot of sysfs_filename:value key:value pairs for this kind of thing.

Quoted 6 lines

Adding those sysfs attributes seems like exactly the thing that people will keep forgetting to do, as there is no (real) functionality depending on them.

I doubt there is any better 'description' of the 'state' of the driver than SHA of the topmost commit + the tree it's related to.

This is exactly what I have been saying in my inward facing roles. Here I'm trying to make sure I'm not missing something that makes this not 100% accurate.

For specific things, I could see those sysfs attributes being useful, but as you say, they are informational only.

Signature · 2 lines
Darren Hart
VMware Open Source Technology Center