Re: xenstored socket backlog length

[Juergen Gross <jgross@xxxxxxxx>]

On 18.07.22 11:01, Roger Pau Monné wrote:
> Hello,
>
> It has been raised on the freebsd-xen mailing list [0] that the socket
> queue length for the xenstored local domain socket is set to 1, which
> can cause concurrent executions of xl commands to fail.
>
> I see in xenstored implementation (xenstored_core.c init_sockets())
> that the call to listen() is made setting a backlog length to 1, and
> hence would like to ask if there's a reasoning for this, as I would
> think having a slightly longer pending connections queue shouldn't be
> an issue.
>
> Was this value chosen based on a toolstack that has a central daemon
> with a single connection to xenstored?
I don't know the reason for the value "1" (it seems to have been like
this since the initial version), maybe it is unlikely on Linux to cause
any problems (there is a note on the listen(2) man-page regarding a
change in the semantics of the backlog parameter in Linux kernel 2.2).

I don't think there will be any problem raising this value to e.g. 16.


Juergen

Attachment: OpenPGP_0xB0DE9DD628BF132F.asc
Description: OpenPGP public key

Attachment: OpenPGP_signature
Description: OpenPGP digital signature