Re: network outages
"Greg A. Woods" <[email protected]> Sat, 21 Mar 2026 15:01:57 -0700
| Newsgroups | gmane.os.netbsd.current |
|---|---|
| Organization | Planix, Inc. |
| Message-ID | <[email protected]> |
At Sat, 21 Mar 2026 16:54:36 -0000 (UTC), [email protected] (Michael van Elst) wrote: Subject: Re: network outages > > 262144 * 2048 / ( 512 + 2048 ) = 209715. > > The relevant setting is kern.sbmax. Setting a socket buffer beyond > sbmax * MCLBYTES / (MSIZE + MCLBYTES) fails and the TCP protocol > no longer attaches. > > The correction factor obviously should make kern.sbmax show the > total memory used for a buffer including the headers. But it's > just confusing. Isn't the simplest to just make kern.sbmax the right size for the largest given {send,recv}buf_max value? I.e.: kern.sbmax = buf_max + ((buf_max / MCLBYTES) * MSIZE) No? Could the kernel do this itself and eliminate the need to manually tune kern.sbmax? Maybe even eliminate kern.sbmax entirely? So, folks using 16Mb for sendbuf_max need to set kern.sbmax to 20Mb, no? -- Greg A. Woods <[email protected]> Kelowna, BC +1 250 762-7675 RoboHack <[email protected]> Planix, Inc. <[email protected]> Avoncote Farms <[email protected]>
signature.asc
(application/pgp-signature, 195 B)
-----BEGIN PGP SIGNATURE----- iF0EABECAB0WIQRuK6dmwVAucmRxuh9mfXG3eL/0fwUCab8VTAAKCRBmfXG3eL/0 f9GYAKDw5jy5jbu+cc4QC8nL0CUfJRwnqwCeJYzoovn0Nw0JMypfNkncmP2/V1U= =zAvw -----END PGP SIGNATURE-----