Return-Path: linux-nfs-owner@vger.kernel.org Received: from aserp1040.oracle.com ([141.146.126.69]:37722 "EHLO aserp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1947043Ab3BHVAP convert rfc822-to-8bit (ORCPT ); Fri, 8 Feb 2013 16:00:15 -0500 Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 (Mac OS X Mail 6.2 \(1499\)) Subject: Re: [PATCH v3 2/2] nfsd: keep a checksum of the first 256 bytes of request From: Chuck Lever In-Reply-To: <20130208205555.GC21040@fieldses.org> Date: Fri, 8 Feb 2013 15:59:53 -0500 Cc: Jeff Layton , linux-nfs@vger.kernel.org, rees@umich.edu Message-Id: References: <1360248701-23963-1-git-send-email-jlayton@redhat.com> <1360248701-23963-3-git-send-email-jlayton@redhat.com> <20130207130316.0cf29993@corrin.poochiereds.net> <20130208082706.614035e2@tlielax.poochiereds.net> <20130208205555.GC21040@fieldses.org> To: "J. Bruce Fields" Sender: linux-nfs-owner@vger.kernel.org List-ID: On Feb 8, 2013, at 3:55 PM, "J. Bruce Fields" wrote: > On Fri, Feb 08, 2013 at 08:27:06AM -0500, Jeff Layton wrote: >> On Thu, 7 Feb 2013 13:03:16 -0500 >> Jeff Layton wrote: >> >>> On Thu, 7 Feb 2013 10:51:02 -0500 >>> Chuck Lever wrote: >>> >>>> >>>> On Feb 7, 2013, at 9:51 AM, Jeff Layton wrote: >>>> >>>>> Now that we're allowing more DRC entries, it becomes a lot easier to hit >>>>> problems with XID collisions. In order to mitigate those, calculate the >>>>> crc32 of up to the first 256 bytes of each request coming in and store >>>>> that in the cache entry, along with the total length of the request. >>>> >>>> I'm happy to see a checksummed DRC finally become reality for the Linux NFS server. >>>> >>>> Have you measured the CPU utilization impact and CPU cache footprint of performing a CRC computation for every incoming RPC? I'm wondering if a simpler checksum might be just as useful but less costly to compute. >>>> >>> >>> No, I haven't, at least not in any sort of rigorous way. It's pretty >>> negligible on "normal" PC hardware, but I think most intel and amd cpus >>> have instructions for handling crc32. I'm ok with a different checksum, >>> we don't need anything cryptographically secure here. I simply chose >>> crc32 since it has an easily available API, and I figured it would be >>> fairly lightweight. >>> >> >> After an abortive attempt to measure this with ftrace, I ended up >> hacking together a patch to just measure the latency of the >> nfsd_cache_csum/_crc functions to get some rough numbers. On my x86_64 >> KVM guest, the avg time to calculate the crc32 is ~1750ns. Using IP >> checksums cuts that roughly in half to ~800ns. I'm not sure how best to >> measure the cache footprint however. >> >> Neither seems terribly significant, especially given the other >> inefficiencies in this code. OTOH, I guess those latencies can add up, >> and I don't see any need to use crc32 over the net/checksum.h routines. >> We probably ought to go with my RFC patch from yesterday. > > OK, I hadn't committed the original yet, so I've just rolled them > together and added a little of the above to the changelog. Look OK? > Chuck, should I add a Reviewed-by: ? Not sure my participation counts as review. How about: Stones-thrown-by: Chuck Lever > --b. > > commit a937bd422ccc4306cdc81b5aa60b12a7212b70d3 > Author: Jeff Layton > Date: Mon Feb 4 11:57:27 2013 -0500 > > nfsd: keep a checksum of the first 256 bytes of request > > Now that we're allowing more DRC entries, it becomes a lot easier to hit > problems with XID collisions. In order to mitigate those, calculate a > checksum of up to the first 256 bytes of each request coming in and store > that in the cache entry, along with the total length of the request. > > This initially used crc32, but Chuck Lever and Jim Rees pointed out that > crc32 is probably more heavyweight than we really need for generating > these checksums, and recommended looking at using the same routines that > are used to generate checksums for IP packets. > > On an x86_64 KVM guest measurements with ftrace showed ~800ns to use > csum_partial vs ~1750ns for crc32. The difference probably isn't > terribly significant, but for now we may as well use csum_partial. > > Signed-off-by: Jeff Layton > Signed-off-by: J. Bruce Fields > > diff --git a/fs/nfsd/cache.h b/fs/nfsd/cache.h > index 9c7232b..87fd141 100644 > --- a/fs/nfsd/cache.h > +++ b/fs/nfsd/cache.h > @@ -29,6 +29,8 @@ struct svc_cacherep { > u32 c_prot; > u32 c_proc; > u32 c_vers; > + unsigned int c_len; > + __wsum c_csum; > unsigned long c_timestamp; > union { > struct kvec u_vec; > @@ -73,6 +75,9 @@ enum { > /* Cache entries expire after this time period */ > #define RC_EXPIRE (120 * HZ) > > +/* Checksum this amount of the request */ > +#define RC_CSUMLEN (256U) > + > int nfsd_reply_cache_init(void); > void nfsd_reply_cache_shutdown(void); > int nfsd_cache_lookup(struct svc_rqst *); > diff --git a/fs/nfsd/nfscache.c b/fs/nfsd/nfscache.c > index f754469..40db57e 100644 > --- a/fs/nfsd/nfscache.c > +++ b/fs/nfsd/nfscache.c > @@ -11,6 +11,7 @@ > #include > #include > #include > +#include > > #include "nfsd.h" > #include "cache.h" > @@ -130,6 +131,7 @@ int nfsd_reply_cache_init(void) > INIT_LIST_HEAD(&lru_head); > max_drc_entries = nfsd_cache_size_limit(); > num_drc_entries = 0; > + > return 0; > out_nomem: > printk(KERN_ERR "nfsd: failed to allocate reply cache\n"); > @@ -238,12 +240,45 @@ nfsd_reply_cache_shrink(struct shrinker *shrink, struct shrink_control *sc) > } > > /* > + * Walk an xdr_buf and get a CRC for at most the first RC_CSUMLEN bytes > + */ > +static __wsum > +nfsd_cache_csum(struct svc_rqst *rqstp) > +{ > + int idx; > + unsigned int base; > + __wsum csum; > + struct xdr_buf *buf = &rqstp->rq_arg; > + const unsigned char *p = buf->head[0].iov_base; > + size_t csum_len = min_t(size_t, buf->head[0].iov_len + buf->page_len, > + RC_CSUMLEN); > + size_t len = min(buf->head[0].iov_len, csum_len); > + > + /* rq_arg.head first */ > + csum = csum_partial(p, len, 0); > + csum_len -= len; > + > + /* Continue into page array */ > + idx = buf->page_base / PAGE_SIZE; > + base = buf->page_base & ~PAGE_MASK; > + while (csum_len) { > + p = page_address(buf->pages[idx]) + base; > + len = min(PAGE_SIZE - base, csum_len); > + csum = csum_partial(p, len, csum); > + csum_len -= len; > + base = 0; > + ++idx; > + } > + return csum; > +} > + > +/* > * Search the request hash for an entry that matches the given rqstp. > * Must be called with cache_lock held. Returns the found entry or > * NULL on failure. > */ > static struct svc_cacherep * > -nfsd_cache_search(struct svc_rqst *rqstp) > +nfsd_cache_search(struct svc_rqst *rqstp, __wsum csum) > { > struct svc_cacherep *rp; > struct hlist_node *hn; > @@ -257,6 +292,7 @@ nfsd_cache_search(struct svc_rqst *rqstp) > hlist_for_each_entry(rp, hn, rh, c_hash) { > if (xid == rp->c_xid && proc == rp->c_proc && > proto == rp->c_prot && vers == rp->c_vers && > + rqstp->rq_arg.len == rp->c_len && csum == rp->c_csum && > rpc_cmp_addr(svc_addr(rqstp), (struct sockaddr *)&rp->c_addr) && > rpc_get_port(svc_addr(rqstp)) == rpc_get_port((struct sockaddr *)&rp->c_addr)) > return rp; > @@ -277,6 +313,7 @@ nfsd_cache_lookup(struct svc_rqst *rqstp) > u32 proto = rqstp->rq_prot, > vers = rqstp->rq_vers, > proc = rqstp->rq_proc; > + __wsum csum; > unsigned long age; > int type = rqstp->rq_cachetype; > int rtn; > @@ -287,10 +324,12 @@ nfsd_cache_lookup(struct svc_rqst *rqstp) > return RC_DOIT; > } > > + csum = nfsd_cache_csum(rqstp); > + > spin_lock(&cache_lock); > rtn = RC_DOIT; > > - rp = nfsd_cache_search(rqstp); > + rp = nfsd_cache_search(rqstp, csum); > if (rp) > goto found_entry; > > @@ -318,7 +357,7 @@ nfsd_cache_lookup(struct svc_rqst *rqstp) > * Must search again just in case someone inserted one > * after we dropped the lock above. > */ > - found = nfsd_cache_search(rqstp); > + found = nfsd_cache_search(rqstp, csum); > if (found) { > nfsd_reply_cache_free_locked(rp); > rp = found; > @@ -344,6 +383,8 @@ setup_entry: > rpc_set_port((struct sockaddr *)&rp->c_addr, rpc_get_port(svc_addr(rqstp))); > rp->c_prot = proto; > rp->c_vers = vers; > + rp->c_len = rqstp->rq_arg.len; > + rp->c_csum = csum; > > hash_refile(rp); > lru_put_end(rp); -- Chuck Lever chuck[dot]lever[at]oracle[dot]com