[Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc?
Anybody have any objections to my removing the ntirpc version? Clearly, this is done in RPCAL/nfs_duprec.c, so why is it also in libntirpc/src/svc_dg.c? According to blame, Matt, Malahal, and Frank have all worked on this, but not since early 2015. -- Announcing the Oxford Dictionaries API! The API offers world-renowned dictionary content that is easy and intuitive to access. Sign up for an account today to start using our lexical data to power your apps and projects. Get started today and enter our developer competition. http://sdm.link/oxford ___ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc?
Hi Bill, As I explained in detail, this code path is used in svc_run, but nfs-ganesha doesn't use it. Matt - Original Message - > From: "William Allen Simpson" > To: "NFS Ganesha Developers" > Sent: Thursday, March 9, 2017 9:21:38 AM > Subject: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc? > > Anybody have any objections to my removing the ntirpc version? > > Clearly, this is done in RPCAL/nfs_duprec.c, so why is it also in > libntirpc/src/svc_dg.c? > > According to blame, Matt, Malahal, and Frank have all worked on this, > but not since early 2015. > > -- > Announcing the Oxford Dictionaries API! The API offers world-renowned > dictionary content that is easy and intuitive to access. Sign up for an > account today to start using our lexical data to power your apps and > projects. Get started today and enter our developer competition. > http://sdm.link/oxford > ___ > Nfs-ganesha-devel mailing list > Nfs-ganesha-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel > -- Matt Benjamin Red Hat, Inc. 315 West Huron Street, Suite 140A Ann Arbor, Michigan 48103 http://www.redhat.com/en/technologies/storage tel. 734-821-5101 fax. 734-769-8938 cel. 734-216-5309 -- Announcing the Oxford Dictionaries API! The API offers world-renowned dictionary content that is easy and intuitive to access. Sign up for an account today to start using our lexical data to power your apps and projects. Get started today and enter our developer competition. http://sdm.link/oxford ___ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc?
It probably should stay. ntirpc is intended to be useful to more than Ganesha, and this seems like a useful feature for potential users. It's not the codepaths called by Ganesha, so it doesn't cause any problems. Daniel On 03/09/2017 09:21 AM, William Allen Simpson wrote: > Anybody have any objections to my removing the ntirpc version? > > Clearly, this is done in RPCAL/nfs_duprec.c, so why is it also in > libntirpc/src/svc_dg.c? > > According to blame, Matt, Malahal, and Frank have all worked on this, > but not since early 2015. > > -- > Announcing the Oxford Dictionaries API! The API offers world-renowned > dictionary content that is easy and intuitive to access. Sign up for an > account today to start using our lexical data to power your apps and > projects. Get started today and enter our developer competition. > http://sdm.link/oxford > ___ > Nfs-ganesha-devel mailing list > Nfs-ganesha-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel > -- Announcing the Oxford Dictionaries API! The API offers world-renowned dictionary content that is easy and intuitive to access. Sign up for an account today to start using our lexical data to power your apps and projects. Get started today and enter our developer competition. http://sdm.link/oxford ___ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc?
On 3/9/17 10:12 AM, Daniel Gryniewicz wrote: > It probably should stay. ntirpc is intended to be useful to more than > Ganesha, and this seems like a useful feature for potential users. It's > not the codepaths called by Ganesha, so it doesn't cause any problems. > It *is* the codepaths called by Ganesha: svc_dg_recv(struct svc_req *req) { [...] req->rq_cksum = #if 1 CityHash64WithSeed(iov.iov_base, MIN(256, iov.iov_len), 103); #else calculate_crc32c(0, iov.iov_base, MIN(256, iov.iov_len)); #endif if (su->su_cache != NULL) { if (svc_dg_cache_get(xprt, &req->rq_msg, &reply, &replylen)) { [...] svc_dg_reply(struct svc_req *req) { [...] if (sendmsg(xprt->xp_fd, msg, 0) == (ssize_t) slen) { stat = true; if (su->su_cache) svc_dg_cache_set(xprt, slen, req->rq_msg.rm_xid); } -- Announcing the Oxford Dictionaries API! The API offers world-renowned dictionary content that is easy and intuitive to access. Sign up for an account today to start using our lexical data to power your apps and projects. Get started today and enter our developer competition. http://sdm.link/oxford ___ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc?
But, isn't su_cache...NULL? Matt - Original Message - > From: "William Allen Simpson" > To: d...@redhat.com, nfs-ganesha-devel@lists.sourceforge.net > Sent: Thursday, March 9, 2017 1:19:21 PM > Subject: Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and > ntirpc? > > On 3/9/17 10:12 AM, Daniel Gryniewicz wrote: > > It probably should stay. ntirpc is intended to be useful to more than > > Ganesha, and this seems like a useful feature for potential users. It's > > not the codepaths called by Ganesha, so it doesn't cause any problems. > > > It *is* the codepaths called by Ganesha: > > svc_dg_recv(struct svc_req *req) > { > [...] > req->rq_cksum = > #if 1 > CityHash64WithSeed(iov.iov_base, MIN(256, iov.iov_len), 103); > #else > calculate_crc32c(0, iov.iov_base, MIN(256, iov.iov_len)); > #endif > > if (su->su_cache != NULL) { > if (svc_dg_cache_get(xprt, &req->rq_msg, &reply, &replylen)) > { > > [...] > > svc_dg_reply(struct svc_req *req) > { > [...] > if (sendmsg(xprt->xp_fd, msg, 0) == (ssize_t) slen) { > stat = true; > if (su->su_cache) > svc_dg_cache_set(xprt, slen, > req->rq_msg.rm_xid); > } > > > -- > Announcing the Oxford Dictionaries API! The API offers world-renowned > dictionary content that is easy and intuitive to access. Sign up for an > account today to start using our lexical data to power your apps and > projects. Get started today and enter our developer competition. > http://sdm.link/oxford > ___ > Nfs-ganesha-devel mailing list > Nfs-ganesha-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel > -- Matt Benjamin Red Hat, Inc. 315 West Huron Street, Suite 140A Ann Arbor, Michigan 48103 http://www.redhat.com/en/technologies/storage tel. 734-821-5101 fax. 734-769-8938 cel. 734-216-5309 -- Announcing the Oxford Dictionaries API! The API offers world-renowned dictionary content that is easy and intuitive to access. Sign up for an account today to start using our lexical data to power your apps and projects. Get started today and enter our developer competition. http://sdm.link/oxford ___ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc?
On 3/9/17 1:44 PM, Matt Benjamin wrote: > But, isn't su_cache...NULL? > Aha, I see that you are correct. It is only set non-NULL in svc_dg_enablecache(), and that's never called. Anywhere. So we have this useless facility that I (and Malahal) have been trying to keep up-to-date with changes, and I've recently fixed the XXX !MT-SAFE (e89139b), and that's all for nought. We don't cache TCP. We don't cache RDMA. This code is an anachronism, and needs to be purged with extreme prejudice. It was badly written, and it's a shame to keep fixing. If we get rid of it, we can use _ioq for output, and get rid of the extra locks. Don't know how important to speed up UDP, but we could -- Announcing the Oxford Dictionaries API! The API offers world-renowned dictionary content that is easy and intuitive to access. Sign up for an account today to start using our lexical data to power your apps and projects. Get started today and enter our developer competition. http://sdm.link/oxford ___ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and ntirpc?
Hi, - Original Message - > From: "William Allen Simpson" > To: "Matt Benjamin" > Cc: d...@redhat.com, nfs-ganesha-devel@lists.sourceforge.net > Sent: Friday, March 10, 2017 2:21:41 PM > Subject: Re: [Nfs-ganesha-devel] UDP duplicate cache in both Ganesha and > ntirpc? > > On 3/9/17 1:44 PM, Matt Benjamin wrote: > > But, isn't su_cache...NULL? > > > Aha, I see that you are correct. It is only set non-NULL in > svc_dg_enablecache(), and that's never called. Anywhere. > > So we have this useless facility that I (and Malahal) have been > trying to keep up-to-date with changes, and I've recently fixed > the XXX !MT-SAFE (e89139b), and that's all for nought. > > We don't cache TCP. We don't cache RDMA. > > This code is an anachronism, and needs to be purged with extreme > prejudice. It was badly written, and it's a shame to keep fixing. Well, it is what it is. Pretty durn old, and never used in nfs-ganesha, I don't believe. > > If we get rid of it, we can use _ioq for output, and get rid of > the extra locks. Don't know how important to speed up UDP, but > we could This seems potentially a useful improvement, I wold say, so that provides some positive motivation for gc'ing the legacy cache stuff, I guess. Matt > > -- Matt Benjamin Red Hat, Inc. 315 West Huron Street, Suite 140A Ann Arbor, Michigan 48103 http://www.redhat.com/en/technologies/storage tel. 734-821-5101 fax. 734-769-8938 cel. 734-216-5309 -- Announcing the Oxford Dictionaries API! The API offers world-renowned dictionary content that is easy and intuitive to access. Sign up for an account today to start using our lexical data to power your apps and projects. Get started today and enter our developer competition. http://sdm.link/oxford ___ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel