Mailing List Archive

udev-208 eats CPU
Hi,

I have just upgraded to the latest build of udev (208) on my Alpha box.
It appears to run at 100% of CPU. When I enable debugging, I get these
lines in /var/log/messages:

unable to receive ctrl connection: Function not implemented

This appears to be due to the lack of the accept4 call.

However, my kernel is gentoo-sources-3.10.7 and my glibc is 2.17,
so I'd expect this call to be there.... Right?

--
Andrew
Re: udev-208 eats CPU [ In reply to ]
A further data point on this issue: downgrading to udev-204 fixes
(i.e. works around) the problem.

So is accept4() really not implemented on the alpha architecture?

--
Andrew
Re: Re: udev-208 eats CPU [ In reply to ]
Andrew Gaylard wrote:
> A further data point on this issue: downgrading to udev-204 fixes
> (i.e. works around) the problem.
>
> So is accept4() really not implemented on the alpha architecture?
>
>
From some quick browsing of
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/ it
looks like a patch was submitted in 2011 and it appeared in kernel 3.2.
Re: udev-208 eats CPU [ In reply to ]
On Sun, Jan 12, 2014 at 12:33 PM, Andrew Gaylard <ag@computer.org> wrote:
> Hi,
>
> I have just upgraded to the latest build of udev (208) on my Alpha box.
> It appears to run at 100% of CPU. When I enable debugging, I get these
> lines in /var/log/messages:
>
> unable to receive ctrl connection: Function not implemented
>
> This appears to be due to the lack of the accept4 call.
>
> However, my kernel is gentoo-sources-3.10.7 and my glibc is 2.17,
> so I'd expect this call to be there.... Right?

Yes, as Alan said accept4 has been supported in the kernel since v3.2.

I'm running udev-208 on a couple of alphas without trouble, so this
problem should be fixable.

Is it possible that your glibc was built against linux-headers older
than v3.2? Seems unlikely, but it's probably the first thing I'd
check.
Re: udev-208 eats CPU [ In reply to ]
On 01/19/14 01:20, Matt Turner wrote:
> On Sun, Jan 12, 2014 at 12:33 PM, Andrew Gaylard <ag@computer.org> wrote:
>> Hi,
>>
>> I have just upgraded to the latest build of udev (208) on my Alpha box.
>> It appears to run at 100% of CPU. When I enable debugging, I get these
>> lines in /var/log/messages:
>>
>> unable to receive ctrl connection: Function not implemented
>>
>> This appears to be due to the lack of the accept4 call.
>>
>> However, my kernel is gentoo-sources-3.10.7 and my glibc is 2.17,
>> so I'd expect this call to be there.... Right?
> Yes, as Alan said accept4 has been supported in the kernel since v3.2.
>
> I'm running udev-208 on a couple of alphas without trouble, so this
> problem should be fixable.
>
> Is it possible that your glibc was built against linux-headers older
> than v3.2? Seems unlikely, but it's probably the first thing I'd
> check.
Yup, you were right. I had sys-kernel/linux-headers 2.6.<something>.
Upgrading that to 3.9 and then rebuilding glibc and udev fixed it.

Thanks!

I'm a bit puzzled why "emerge -uva @system" and "emerge -uva @world"
didn't automatically upgrade linux-headers. I had to explicitly request it.
But hey, it's working now...:)

--
Andrew Gaylard
Re: udev-208 eats CPU [ In reply to ]
Andrew Gaylard <ag@computer.org> writes:

> On 01/19/14 01:20, Matt Turner wrote:
>> On Sun, Jan 12, 2014 at 12:33 PM, Andrew Gaylard <ag@computer.org> wrote:
>>> Hi,
>>>
>>> I have just upgraded to the latest build of udev (208) on my Alpha box.
>>> It appears to run at 100% of CPU. When I enable debugging, I get these
>>> lines in /var/log/messages:
>>>
>>> unable to receive ctrl connection: Function not implemented
>>>
>>> This appears to be due to the lack of the accept4 call.
>>>
>>> However, my kernel is gentoo-sources-3.10.7 and my glibc is 2.17,
>>> so I'd expect this call to be there.... Right?
>> Yes, as Alan said accept4 has been supported in the kernel since v3.2.
>>
>> I'm running udev-208 on a couple of alphas without trouble, so this
>> problem should be fixable.
>>
>> Is it possible that your glibc was built against linux-headers older
>> than v3.2? Seems unlikely, but it's probably the first thing I'd
>> check.
> Yup, you were right. I had sys-kernel/linux-headers 2.6.<something>.
> Upgrading that to 3.9 and then rebuilding glibc and udev fixed it.
>
> Thanks!
>
> I'm a bit puzzled why "emerge -uva @system" and "emerge -uva @world"
> didn't automatically upgrade linux-headers. I had to explicitly request it.
> But hey, it's working now...:)

Try adding -D to those flags. That usually does the trick.

--
Måns Rullgård
mans@mansr.com
Re: Re: udev-208 eats CPU [ In reply to ]
On 02/12/14 22:15, Måns Rullgård wrote:
> Andrew Gaylard <ag@computer.org> writes:
>
>> On 01/19/14 01:20, Matt Turner wrote:
>>> On Sun, Jan 12, 2014 at 12:33 PM, Andrew Gaylard <ag@computer.org> wrote:
>>>> Hi,
>>>>
>>>> I have just upgraded to the latest build of udev (208) on my Alpha box.
>>>> It appears to run at 100% of CPU. When I enable debugging, I get these
>>>> lines in /var/log/messages:
>>>>
>>>> unable to receive ctrl connection: Function not implemented
>>>>
>>>> This appears to be due to the lack of the accept4 call.
>>>>
>>>> However, my kernel is gentoo-sources-3.10.7 and my glibc is 2.17,
>>>> so I'd expect this call to be there.... Right?
>>> Yes, as Alan said accept4 has been supported in the kernel since v3.2.
>>>
>>> I'm running udev-208 on a couple of alphas without trouble, so this
>>> problem should be fixable.
>>>
>>> Is it possible that your glibc was built against linux-headers older
>>> than v3.2? Seems unlikely, but it's probably the first thing I'd
>>> check.
>> Yup, you were right. I had sys-kernel/linux-headers 2.6.<something>.
>> Upgrading that to 3.9 and then rebuilding glibc and udev fixed it.
>>
>> Thanks!
>>
>> I'm a bit puzzled why "emerge -uva @system" and "emerge -uva @world"
>> didn't automatically upgrade linux-headers. I had to explicitly request it.
>> But hey, it's working now...:)
> Try adding -D to those flags. That usually does the trick.
>
Oh wow. This box was installed with Gentoo back in 2005.
There's a *lot* to catch up! This'll build for a while...:)

Thanks a lot for the tip -- it definitely does the trick.

--
Andrew Gaylard