Re: [squid-dev] [PATCH] Do not make bogus recvmsg calls
On 06/09/2016 08:09 PM, Amos Jeffries wrote: > On 10/06/2016 12:20 p.m., Alex Rousskov wrote: >> The attached simple patch prevents bogus recvmsg(2) calls when >> closing file descriptors. Please see the patch preamble for details. > +1. Looks okay for now. Committed to trunk (r14714). > Its relation to RST implies UDP sockets are also doing useless reads. > IIRC there is a difference between client and server TCP sockets RST > behaviour as well. So this probably needs a full re-analysis to get right. Agreed. Thank you, Alex. ___ squid-dev mailing list squid-dev@lists.squid-cache.org http://lists.squid-cache.org/listinfo/squid-dev
Re: [squid-dev] [PATCH] Do not make bogus recvmsg calls
On 10/06/2016 12:20 p.m., Alex Rousskov wrote: > Hello, > > The attached simple patch prevents bogus recvmsg(2) calls when > closing file descriptors. Please see the patch preamble for details. > > A cleaner way to handle this problem may be to introduce a new > FD_FLUSH_METHOD() API that would do nothing for UDS (and possibly BIO > SSL?) sockets. However, a whole new API feels like a too-heavy solution > for such a small/isolated/hacky purpose. It may be better to spend that > development cycles on avoiding flushing TCP sockets that already > returned a zero-size read and do not need to be flushed. Please let me > know if you disagree (and I will not commit this patch). > +1. Looks okay for now. Its relation to RST implies UDP sockets are also doing useless reads. IIRC there is a difference between client and server TCP sockets RST behaviour as well. So this probably needs a full re-analysis to get right. Amos ___ squid-dev mailing list squid-dev@lists.squid-cache.org http://lists.squid-cache.org/listinfo/squid-dev