2014-02-20 17:44:53

by Michael Opdenacker

[permalink] [raw]
Subject: How to get rid of IRQF_DISABLED for good?

Hi,

In spite of the patches I have been sending (and resending!) over the
past months, there are still 118 occurrences of the idle IRQF_DISABLED
flag in the kernel code. This corresponds to 31 patches which haven't
been accepted yet.

What would you advise to get rid of IRQF_DISABLED for good?

* Send a treewide patch removing the last occurrences in one shot,
bypassing the regular maintainers? Who could take it?
* Remove the definition of IRQF_DISABLED to force the individual
maintainers (and out of tree drivers!) to update their code? It
could be a way of seeing which code isn't maintained any more ;)
* Continue to resend the patches for a few more cycles, until the
corresponding maintainers can no longer bear the discredit?
* Any other solution?

Thank you in advance for your advise!

Michael.

--
Michael Opdenacker, CEO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
+33 484 258 098


2014-02-20 18:18:00

by Levente Kurusa

[permalink] [raw]
Subject: Re: How to get rid of IRQF_DISABLED for good?

2014-02-20 18:44 GMT+01:00 Michael Opdenacker
<[email protected]>:
> Hi,
>
> In spite of the patches I have been sending (and resending!) over the
> past months, there are still 118 occurrences of the idle IRQF_DISABLED
> flag in the kernel code. This corresponds to 31 patches which haven't
> been accepted yet.
>
> What would you advise to get rid of IRQF_DISABLED for good?
>
> * Send a treewide patch removing the last occurrences in one shot,
> bypassing the regular maintainers? Who could take it?

Andrew Morton would take it to his -mm tree.
This, IMO, seems to be the best solution to circumvent unresponsive/uncaring
maintainers.

> * Remove the definition of IRQF_DISABLED to force the individual
> maintainers (and out of tree drivers!) to update their code? It
> could be a way of seeing which code isn't maintained any more ;)

No, every single patch has to be 'bisectable' meaning that when you bisect
you should be able to build every single patch as is.

> * Continue to resend the patches for a few more cycles, until the
> corresponding maintainers can no longer bear the discredit?

Maybe once more, if they don't reply, send it to Andrew Morton as well
and CC a few people who know your work is good so that they can ACK it.

Oh and maybe you could add an __attribute__((deprecated)) to it, but
I am not sure that's possible and/or correct.


--
Regards,
Levente Kurusa