/*      $NetBSD: nfs_fha.c,v 1.2 2016/12/13 22:41:46 pgoyette Exp $     */
/*-
* Copyright (c) 2008 Isilon Inc http://www.isilon.com/
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/

#include <sys/cdefs.h>
/* __FBSDID("FreeBSD: head/sys/nfs/nfs_fha.c 267479 2014-06-14 12:26:12Z mav "); */
__RCSID("$NetBSD: nfs_fha.c,v 1.2 2016/12/13 22:41:46 pgoyette Exp $");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sysproto.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/vnode.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mbuf.h>
#include <sys/sbuf.h>

#include <rpc/rpc.h>
#include <fs/nfs/common/nfs_fha.h>

static MALLOC_DEFINE(M_NFS_FHA, "NFS FHA", "NFS FHA");

/*
* XXX need to commonize definitions between old and new NFS code.  Define
* this here so we don't include one nfsproto.h over the other.
*/
#define NFS_PROG                100003

void
fha_init(struct fha_params *softc)
{
       char tmpstr[128];
       int i;

       for (i = 0; i < FHA_HASH_SIZE; i++)
               mtx_init(&softc->fha_hash[i].mtx, "fhalock", NULL, MTX_DEF);

       /*
        * Set the default tuning parameters.
        */
       softc->ctls.enable = FHA_DEF_ENABLE;
       softc->ctls.bin_shift = FHA_DEF_BIN_SHIFT;
       softc->ctls.max_nfsds_per_fh = FHA_DEF_MAX_NFSDS_PER_FH;
       softc->ctls.max_reqs_per_nfsd = FHA_DEF_MAX_REQS_PER_NFSD;

       /*
        * Allow the user to override the defaults at boot time with
        * tunables.
        */
       snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.enable",
           softc->server_name);
       TUNABLE_INT_FETCH(tmpstr, &softc->ctls.enable);
       snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.bin_shift",
           softc->server_name);
       TUNABLE_INT_FETCH(tmpstr, &softc->ctls.bin_shift);
       snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.max_nfsds_per_fh",
           softc->server_name);
       TUNABLE_INT_FETCH(tmpstr, &softc->ctls.max_nfsds_per_fh);
       snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.max_reqs_per_nfsd",
           softc->server_name);
       TUNABLE_INT_FETCH(tmpstr, &softc->ctls.max_reqs_per_nfsd);

       /*
        * Add sysctls so the user can change the tuning parameters at
        * runtime.
        */
       SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
           OID_AUTO, "enable", CTLFLAG_RW,
           &softc->ctls.enable, 0, "Enable NFS File Handle Affinity (FHA)");

       SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
           OID_AUTO, "bin_shift", CTLFLAG_RW,
           &softc->ctls.bin_shift, 0, "For FHA reads, no two requests will "
           "contend if they're 2^(bin_shift) bytes apart");

       SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
           OID_AUTO, "max_nfsds_per_fh", CTLFLAG_RW,
           &softc->ctls.max_nfsds_per_fh, 0, "Maximum nfsd threads that "
           "should be working on requests for the same file handle");

       SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
           OID_AUTO, "max_reqs_per_nfsd", CTLFLAG_RW,
           &softc->ctls.max_reqs_per_nfsd, 0, "Maximum requests that "
           "single nfsd thread should be working on at any time");

       SYSCTL_ADD_OID(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
           OID_AUTO, "fhe_stats", CTLTYPE_STRING | CTLFLAG_RD, 0, 0,
           softc->callbacks.fhe_stats_sysctl, "A", "");

}

void
fha_uninit(struct fha_params *softc)
{
       int i;

       sysctl_ctx_free(&softc->sysctl_ctx);
       for (i = 0; i < FHA_HASH_SIZE; i++)
               mtx_destroy(&softc->fha_hash[i].mtx);
}

/*
* This just specifies that offsets should obey affinity when within
* the same 1Mbyte (1<<20) chunk for the file (reads only for now).
*/
static void
fha_extract_info(struct svc_req *req, struct fha_info *i,
   struct fha_callbacks *cb)
{
       struct mbuf *md;
       caddr_t dpos;
       static u_int64_t random_fh = 0;
       int error;
       int v3 = (req->rq_vers == 3);
       rpcproc_t procnum;

       /*
        * We start off with a random fh.  If we get a reasonable
        * procnum, we set the fh.  If there's a concept of offset
        * that we're interested in, we set that.
        */
       i->fh = ++random_fh;
       i->offset = 0;
       i->locktype = LK_EXCLUSIVE;

       /*
        * Extract the procnum and convert to v3 form if necessary,
        * taking care to deal with out-of-range procnums.  Caller will
        * ensure that rq_vers is either 2 or 3.
        */
       procnum = req->rq_proc;
       if (!v3) {
               rpcproc_t tmp_procnum;

               tmp_procnum = cb->get_procnum(procnum);
               if (tmp_procnum == -1)
                       goto out;
               procnum = tmp_procnum;
       }

       /*
        * We do affinity for most.  However, we divide a realm of affinity
        * by file offset so as to allow for concurrent random access.  We
        * only do this for reads today, but this may change when IFS supports
        * efficient concurrent writes.
        */
       if (cb->no_offset(procnum))
               goto out;

       error = cb->realign(&req->rq_args, M_NOWAIT);
       if (error)
               goto out;
       md = req->rq_args;
       dpos = mtod(md, caddr_t);

       /* Grab the filehandle. */
       error = cb->get_fh(&i->fh, v3, &md, &dpos);
       if (error)
               goto out;

       /* Content ourselves with zero offset for all but reads. */
       if (cb->is_read(procnum) || cb->is_write(procnum))
               cb->get_offset(&md, &dpos, v3, i);

out:
       cb->set_locktype(procnum, i);
}

static struct fha_hash_entry *
fha_hash_entry_new(u_int64_t fh)
{
       struct fha_hash_entry *e;

       e = malloc(sizeof(*e), M_NFS_FHA, M_WAITOK);
       e->fh = fh;
       e->num_rw = 0;
       e->num_exclusive = 0;
       e->num_threads = 0;
       LIST_INIT(&e->threads);

       return (e);
}

static void
fha_hash_entry_destroy(struct fha_hash_entry *e)
{

       mtx_assert(e->mtx, MA_OWNED);
       KASSERT(e->num_rw == 0,
           ("%d reqs on destroyed fhe %p", e->num_rw, e));
       KASSERT(e->num_exclusive == 0,
           ("%d exclusive reqs on destroyed fhe %p", e->num_exclusive, e));
       KASSERT(e->num_threads == 0,
           ("%d threads on destroyed fhe %p", e->num_threads, e));
       free(e, M_NFS_FHA);
}

static void
fha_hash_entry_remove(struct fha_hash_entry *e)
{

       mtx_assert(e->mtx, MA_OWNED);
       LIST_REMOVE(e, link);
       fha_hash_entry_destroy(e);
}

static struct fha_hash_entry *
fha_hash_entry_lookup(struct fha_params *softc, u_int64_t fh)
{
       SVCPOOL *pool;
       struct fha_hash_slot *fhs;
       struct fha_hash_entry *fhe, *new_fhe;

       pool = *softc->pool;
       fhs = &softc->fha_hash[fh % FHA_HASH_SIZE];
       new_fhe = fha_hash_entry_new(fh);
       new_fhe->mtx = &fhs->mtx;
       mtx_lock(&fhs->mtx);
       LIST_FOREACH(fhe, &fhs->list, link)
               if (fhe->fh == fh)
                       break;
       if (!fhe) {
               fhe = new_fhe;
               LIST_INSERT_HEAD(&fhs->list, fhe, link);
       } else
               fha_hash_entry_destroy(new_fhe);
       return (fhe);
}

static void
fha_hash_entry_add_thread(struct fha_hash_entry *fhe, SVCTHREAD *thread)
{

       mtx_assert(fhe->mtx, MA_OWNED);
       thread->st_p2 = 0;
       LIST_INSERT_HEAD(&fhe->threads, thread, st_alink);
       fhe->num_threads++;
}

static void
fha_hash_entry_remove_thread(struct fha_hash_entry *fhe, SVCTHREAD *thread)
{

       mtx_assert(fhe->mtx, MA_OWNED);
       KASSERT(thread->st_p2 == 0,
           ("%d reqs on removed thread %p", thread->st_p2, thread));
       LIST_REMOVE(thread, st_alink);
       fhe->num_threads--;
}

/*
* Account for an ongoing operation associated with this file.
*/
static void
fha_hash_entry_add_op(struct fha_hash_entry *fhe, int locktype, int count)
{

       mtx_assert(fhe->mtx, MA_OWNED);
       if (LK_EXCLUSIVE == locktype)
               fhe->num_exclusive += count;
       else
               fhe->num_rw += count;
}

/*
* Get the service thread currently associated with the fhe that is
* appropriate to handle this operation.
*/
static SVCTHREAD *
fha_hash_entry_choose_thread(struct fha_params *softc,
   struct fha_hash_entry *fhe, struct fha_info *i, SVCTHREAD *this_thread)
{
       SVCTHREAD *thread, *min_thread = NULL;
       SVCPOOL *pool;
       int req_count, min_count = 0;
       off_t offset1, offset2;

       pool = *softc->pool;

       LIST_FOREACH(thread, &fhe->threads, st_alink) {
               req_count = thread->st_p2;

               /* If there are any writes in progress, use the first thread. */
               if (fhe->num_exclusive) {
#if 0
                       ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
                           "fha: %p(%d)w", thread, req_count);
#endif
                       return (thread);
               }

               /*
                * Check for read locality, making sure that we won't
                * exceed our per-thread load limit in the process.
                */
               offset1 = i->offset;
               offset2 = thread->st_p3;

               if (((offset1 >= offset2)
                 && ((offset1 - offset2) < (1 << softc->ctls.bin_shift)))
                || ((offset2 > offset1)
                 && ((offset2 - offset1) < (1 << softc->ctls.bin_shift)))) {
                       if ((softc->ctls.max_reqs_per_nfsd == 0) ||
                           (req_count < softc->ctls.max_reqs_per_nfsd)) {
#if 0
                               ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
                                   "fha: %p(%d)r", thread, req_count);
#endif
                               return (thread);
                       }
               }

               /*
                * We don't have a locality match, so skip this thread,
                * but keep track of the most attractive thread in case
                * we need to come back to it later.
                */
#if 0
               ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
                   "fha: %p(%d)s off1 %llu off2 %llu", thread,
                   req_count, offset1, offset2);
#endif
               if ((min_thread == NULL) || (req_count < min_count)) {
                       min_count = req_count;
                       min_thread = thread;
               }
       }

       /*
        * We didn't find a good match yet.  See if we can add
        * a new thread to this file handle entry's thread list.
        */
       if ((softc->ctls.max_nfsds_per_fh == 0) ||
           (fhe->num_threads < softc->ctls.max_nfsds_per_fh)) {
               thread = this_thread;
#if 0
               ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
                   "fha: %p(%d)t", thread, thread->st_p2);
#endif
               fha_hash_entry_add_thread(fhe, thread);
       } else {
               /*
                * We don't want to use any more threads for this file, so
                * go back to the most attractive nfsd we're already using.
                */
               thread = min_thread;
       }

       return (thread);
}

/*
* After getting a request, try to assign it to some thread.  Usually we
* handle it ourselves.
*/
SVCTHREAD *
fha_assign(SVCTHREAD *this_thread, struct svc_req *req,
   struct fha_params *softc)
{
       SVCTHREAD *thread;
       struct fha_info i;
       struct fha_hash_entry *fhe;
       struct fha_callbacks *cb;

       cb = &softc->callbacks;

       /* Check to see whether we're enabled. */
       if (softc->ctls.enable == 0)
               goto thist;

       /*
        * Only do placement if this is an NFS request.
        */
       if (req->rq_prog != NFS_PROG)
               goto thist;

       if (req->rq_vers != 2 && req->rq_vers != 3)
               goto thist;

       fha_extract_info(req, &i, cb);

       /*
        * We save the offset associated with this request for later
        * nfsd matching.
        */
       fhe = fha_hash_entry_lookup(softc, i.fh);
       req->rq_p1 = fhe;
       req->rq_p2 = i.locktype;
       req->rq_p3 = i.offset;

       /*
        * Choose a thread, taking into consideration locality, thread load,
        * and the number of threads already working on this file.
        */
       thread = fha_hash_entry_choose_thread(softc, fhe, &i, this_thread);
       KASSERT(thread, ("fha_assign: NULL thread!"));
       fha_hash_entry_add_op(fhe, i.locktype, 1);
       thread->st_p2++;
       thread->st_p3 = i.offset;

       /*
        * Grab the pool lock here to not let chosen thread go away before
        * the new request inserted to its queue while we drop fhe lock.
        */
       mtx_lock(&thread->st_lock);
       mtx_unlock(fhe->mtx);

       return (thread);
thist:
       req->rq_p1 = NULL;
       mtx_lock(&this_thread->st_lock);
       return (this_thread);
}

/*
* Called when we're done with an operation.  The request has already
* been de-queued.
*/
void
fha_nd_complete(SVCTHREAD *thread, struct svc_req *req)
{
       struct fha_hash_entry *fhe = req->rq_p1;
       struct mtx *mtx;

       /*
        * This may be called for reqs that didn't go through
        * fha_assign (e.g. extra NULL ops used for RPCSEC_GSS.
        */
       if (!fhe)
               return;

       mtx = fhe->mtx;
       mtx_lock(mtx);
       fha_hash_entry_add_op(fhe, req->rq_p2, -1);
       thread->st_p2--;
       KASSERT(thread->st_p2 >= 0, ("Negative request count %d on %p",
           thread->st_p2, thread));
       if (thread->st_p2 == 0) {
               fha_hash_entry_remove_thread(fhe, thread);
               if (0 == fhe->num_rw + fhe->num_exclusive)
                       fha_hash_entry_remove(fhe);
       }
       mtx_unlock(mtx);
}

int
fhe_stats_sysctl(SYSCTL_HANDLER_ARGS, struct fha_params *softc)
{
       int error, i;
       struct sbuf sb;
       struct fha_hash_entry *fhe;
       bool_t first, hfirst;
       SVCTHREAD *thread;
       SVCPOOL *pool;

       sbuf_new(&sb, NULL, 65536, SBUF_FIXEDLEN);

       pool = NULL;

       if (!*softc->pool) {
               sbuf_printf(&sb, "NFSD not running\n");
               goto out;
       }
       pool = *softc->pool;

       for (i = 0; i < FHA_HASH_SIZE; i++)
               if (!LIST_EMPTY(&softc->fha_hash[i].list))
                       break;

       if (i == FHA_HASH_SIZE) {
               sbuf_printf(&sb, "No file handle entries.\n");
               goto out;
       }

       hfirst = TRUE;
       for (; i < FHA_HASH_SIZE; i++) {
               mtx_lock(&softc->fha_hash[i].mtx);
               if (LIST_EMPTY(&softc->fha_hash[i].list)) {
                       mtx_unlock(&softc->fha_hash[i].mtx);
                       continue;
               }
               sbuf_printf(&sb, "%shash %d: {\n", hfirst ? "" : ", ", i);
               first = TRUE;
               LIST_FOREACH(fhe, &softc->fha_hash[i].list, link) {
                       sbuf_printf(&sb, "%sfhe %p: {\n", first ? "  " : ", ", fhe);

                       sbuf_printf(&sb, "    fh: %ju\n", (uintmax_t) fhe->fh);
                       sbuf_printf(&sb, "    num_rw/exclusive: %d/%d\n",
                           fhe->num_rw, fhe->num_exclusive);
                       sbuf_printf(&sb, "    num_threads: %d\n", fhe->num_threads);

                       LIST_FOREACH(thread, &fhe->threads, st_alink) {
                               sbuf_printf(&sb, "      thread %p offset %ju "
                                   "reqs %d\n", thread,
                                   thread->st_p3, thread->st_p2);
                       }

                       sbuf_printf(&sb, "  }");
                       first = FALSE;
               }
               sbuf_printf(&sb, "\n}");
               mtx_unlock(&softc->fha_hash[i].mtx);
               hfirst = FALSE;
       }

out:
       sbuf_trim(&sb);
       sbuf_finish(&sb);
       error = sysctl_handle_string(oidp, sbuf_data(&sb), sbuf_len(&sb), req);
       sbuf_delete(&sb);
       return (error);
}

You are viewing proxied material from ftp.icm.edu.pl. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.