Hi,
The arp(7) page's description of gc_stale_time doesn't quite describe the behaviour correctly, at least as I understand it.
The current description suggests this is the time interval at which a loop will look for stale entries. However, this field is the threshold for marking an entry dead for removal, based on when it was last used (see net/core/neighbour.c lines 935-942) and whether the table is over gc_thresh1. How often this check is done appears to be determined by base_reachable_time (/2) and the third option gc_interval is not involved in this process as far as I can tell, despite its name.
Perhaps a draft alternate description could be something along the lines of:
Determines the threshold for removing a cache entry after it was last used and when the cache is larger than gc_thresh1. Defaults to 60 seconds.
Thanks,
Adam
Hi,
Any opinion on this?
Adam
On Tue, 16 Feb 2021, at 17:00, Adam Liddell wrote:
Hi,
The arp(7) page's description of gc_stale_time doesn't quite describe
the behaviour correctly, at least as I understand it.
The current description suggests this is the time interval at which a
loop will look for stale entries. However, this field is the threshold
for marking an entry dead for removal, based on when it was last used
(see net/core/neighbour.c lines 935-942) and whether the table is over
gc_thresh1. How often this check is done appears to be determined by
base_reachable_time (/2) and the third option gc_interval is not
involved in this process as far as I can tell, despite its name.
Perhaps a draft alternate description could be something along the lines of:
Determines the threshold for removing a cache entry after it was last
used and when the cache is larger than gc_thresh1. Defaults to 60
seconds.
Thanks,
Adam
[CC += netdev]
On 4/12/21 12:13 AM, Adam Liddell wrote:
Hi,
Any opinion on this?
Adam
On Tue, 16 Feb 2021, at 17:00, Adam Liddell wrote:
quoted
Hi,
The arp(7) page's description of gc_stale_time doesn't quite describe
the behaviour correctly, at least as I understand it.
The current description suggests this is the time interval at which a
loop will look for stale entries. However, this field is the threshold
for marking an entry dead for removal, based on when it was last used
(see net/core/neighbour.c lines 935-942) and whether the table is over
gc_thresh1. How often this check is done appears to be determined by
base_reachable_time (/2) and the third option gc_interval is not
involved in this process as far as I can tell, despite its name.
Perhaps a draft alternate description could be something along the lines of:
Determines the threshold for removing a cache entry after it was last
used and when the cache is larger than gc_thresh1. Defaults to 60
seconds.
Hi Adam,
Thanks for the ping. I don't know, sorry.
I added netdev@ so that maybe they can give an opinion. In the
man-pages, that text has been there since we use git, so I don't know
who wrote it.
Thanks,
Alex
--
Alejandro Colomar
Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
http://www.alejandro-colomar.es/