strerror for 524

strerror for 524

[Jeffrey Walton]

Hi Everyone,

I was testing the code at
https://www.kernel.org/doc/html/latest/networking/tls.html#kernel-tls.
The code returns error 524 for setsockopt. Calling strerror(errno)
results in "Unknown error 524".

Error code 524 seems to be pretty prevalent. [0,1,2,3]] (and friends)

I think the error string for code 524 could be improved. There is a
small list of similar error codes at
https://bugzilla.redhat.com/show_bug.cgi?id=1778348#c1 .

Would glibc mind providing a useful error string.

[0] https://bugzilla.redhat.com/show_bug.cgi?id=1778348
[1] https://patchwork.ozlabs.org/patch/309627/
[2] https://lkml.org/lkml/2019/5/16/883
[3] https://duckduckgo.com/?q="Unknown+error+524"

Re: strerror for 524

[Florian Weimer]

* Jeffrey Walton:

> I was testing the code at
> https://www.kernel.org/doc/html/latest/networking/tls.html#kernel-tls.
> The code returns error 524 for setsockopt. Calling strerror(errno)
> results in "Unknown error 524".
>
> Error code 524 seems to be pretty prevalent. [0,1,2,3]] (and friends)
>
> I think the error string for code 524 could be improved. There is a
> small list of similar error codes at
> https://bugzilla.redhat.com/show_bug.cgi?id=1778348#c1 .
>
> Would glibc mind providing a useful error string.

The error code is not listed in a UAPI header.  Isn't it a kernel bug
that it leaks to userspace?

The internal kernel name for 524 appears to be ENOTSUPP, but the UAPI
headers currently define ENOTSUPP as EOPNOTSUPP, which is 95.  If the
intent is to expose the error code to user space, that would have to
be changed, too.

Re: strerror for 524

[Florian Weimer]

* Florian Weimer:

> * Jeffrey Walton:
>
>> I was testing the code at
>> https://www.kernel.org/doc/html/latest/networking/tls.html#kernel-tls.
>> The code returns error 524 for setsockopt. Calling strerror(errno)
>> results in "Unknown error 524".
>>
>> Error code 524 seems to be pretty prevalent. [0,1,2,3]] (and friends)
>>
>> I think the error string for code 524 could be improved. There is a
>> small list of similar error codes at
>> https://bugzilla.redhat.com/show_bug.cgi?id=1778348#c1 .
>>
>> Would glibc mind providing a useful error string.
>
> The error code is not listed in a UAPI header.  Isn't it a kernel bug
> that it leaks to userspace?
>
> The internal kernel name for 524 appears to be ENOTSUPP, but the UAPI
> headers currently define ENOTSUPP as EOPNOTSUPP, which is 95.  If the
> intent is to expose the error code to user space, that would have to
> be changed, too.

Correction: The UAPI headers do not define ENOTSUPP.  The glibc
headers define ENOTSUP (sic) as an alias for EOPNOTSUPP.

This has already been posted to the kernelnewbies list:

<http://lists.kernelnewbies.org/pipermail/kernelnewbies/2019-November/020483.html>

It should eventually go to linux-crypto and netdev, I assume.