Return-Path: Received: from mail-bw0-f46.google.com ([209.85.214.46]:44630 "EHLO mail-bw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751184Ab0KKPJa convert rfc822-to-8bit (ORCPT ); Thu, 11 Nov 2010 10:09:30 -0500 Received: by bwz15 with SMTP id 15so2029424bwz.19 for ; Thu, 11 Nov 2010 07:09:28 -0800 (PST) In-Reply-To: <4CDC0500.90805@panasas.com> References: <1288884151-11128-1-git-send-email-iisaman@netapp.com> <1288884151-11128-5-git-send-email-iisaman@netapp.com> <4CDC0500.90805@panasas.com> Date: Thu, 11 Nov 2010 10:09:28 -0500 Message-ID: Subject: Re: [PATCH 04/18] pnfs-submit: change layout state seqlock to a spinlock From: Fred Isaman To: Benny Halevy Cc: linux-nfs@vger.kernel.org Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 On Thu, Nov 11, 2010 at 10:00 AM, Benny Halevy wrote: > On 2010-11-04 17:22, Fred Isaman wrote: >> This prepares for future changes, where the layout state needs >> to change atomically with several other variables. ?In particular, >> it will need to know if lo->segs is empty. ?Moreover, the >> layoutstateid is not really a read-mostly structure, as it is >> written on each LAYOUTGET. >> >> Signed-off-by: Fred Isaman >> --- >> ?fs/nfs/callback_proc.c | ? ?8 +++--- >> ?fs/nfs/nfs4xdr.c ? ? ? | ? ?2 + >> ?fs/nfs/pnfs.c ? ? ? ? ?| ? 55 ++++++++++++++--------------------------------- >> ?fs/nfs/pnfs.h ? ? ? ? ?| ? ?4 +-- >> ?4 files changed, 24 insertions(+), 45 deletions(-) >> >> diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c >> index 84c5a1b..3e022a8 100644 >> --- a/fs/nfs/callback_proc.c >> +++ b/fs/nfs/callback_proc.c >> @@ -135,12 +135,11 @@ static bool >> ?pnfs_is_next_layout_stateid(const struct pnfs_layout_hdr *lo, >> ? ? ? ? ? ? ? ? ? ? ? ? ? const nfs4_stateid stateid) >> ?{ >> - ? ? int seqlock; >> ? ? ? bool res; >> ? ? ? u32 oldseqid, newseqid; >> >> - ? ? do { >> - ? ? ? ? ? ? seqlock = read_seqbegin(&lo->seqlock); >> + ? ? spin_lock(&lo->inode->i_lock); >> + ? ? { >> ? ? ? ? ? ? ? oldseqid = be32_to_cpu(lo->stateid.stateid.seqid); >> ? ? ? ? ? ? ? newseqid = be32_to_cpu(stateid.stateid.seqid); >> ? ? ? ? ? ? ? res = !memcmp(lo->stateid.stateid.other, >> @@ -158,7 +157,8 @@ pnfs_is_next_layout_stateid(const struct pnfs_layout_hdr *lo, >> ? ? ? ? ? ? ? ? ? ? ? if (res) >> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? res = (newseqid == 1); >> ? ? ? ? ? ? ? } >> - ? ? } while (read_seqretry(&lo->seqlock, seqlock)); >> + ? ? } >> + ? ? spin_unlock(&lo->inode->i_lock); >> >> ? ? ? return res; >> ?} >> diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c >> index 6d9ef2b..b71a482 100644 >> --- a/fs/nfs/nfs4xdr.c >> +++ b/fs/nfs/nfs4xdr.c >> @@ -1915,7 +1915,9 @@ encode_layoutreturn(struct xdr_stream *xdr, >> ? ? ? ? ? ? ? p = reserve_space(xdr, 16 + NFS4_STATEID_SIZE); >> ? ? ? ? ? ? ? p = xdr_encode_hyper(p, args->range.offset); >> ? ? ? ? ? ? ? p = xdr_encode_hyper(p, args->range.length); >> + ? ? ? ? ? ? spin_lock(&args->inode->i_lock); >> ? ? ? ? ? ? ? pnfs_copy_layout_stateid(&stateid, NFS_I(args->inode)->layout); >> + ? ? ? ? ? ? spin_unlock(&args->inode->i_lock); >> ? ? ? ? ? ? ? p = xdr_encode_opaque_fixed(p, &stateid.data, >> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? NFS4_STATEID_SIZE); >> ? ? ? ? ? ? ? p = reserve_space(xdr, 4); >> diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c >> index 4e5c68b..01ecb95 100644 >> --- a/fs/nfs/pnfs.c >> +++ b/fs/nfs/pnfs.c >> @@ -456,7 +456,7 @@ pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, >> ? ? ? nfs4_stateid *old = &lo->stateid; >> ? ? ? bool overwrite = false; >> >> - ? ? write_seqlock(&lo->seqlock); >> + ? ? assert_spin_locked(&lo->inode->i_lock); >> ? ? ? if (!test_bit(NFS_LAYOUT_STATEID_SET, &lo->state) || >> ? ? ? ? ? memcmp(old->stateid.other, new->stateid.other, sizeof(new->stateid.other))) >> ? ? ? ? ? ? ? overwrite = true; >> @@ -470,54 +470,34 @@ pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, >> ? ? ? } >> ? ? ? if (overwrite) >> ? ? ? ? ? ? ? memcpy(&old->stateid, &new->stateid, sizeof(new->stateid)); >> - ? ? write_sequnlock(&lo->seqlock); >> -} >> - >> -static void >> -pnfs_layout_from_open_stateid(struct pnfs_layout_hdr *lo, >> - ? ? ? ? ? ? ? ? ? ? ? ? ? struct nfs4_state *state) >> -{ >> - ? ? int seq; >> - >> - ? ? dprintk("--> %s\n", __func__); >> - ? ? write_seqlock(&lo->seqlock); >> - ? ? do { >> - ? ? ? ? ? ? seq = read_seqbegin(&state->seqlock); >> - ? ? ? ? ? ? memcpy(lo->stateid.data, state->stateid.data, >> - ? ? ? ? ? ? ? ? ? ?sizeof(state->stateid.data)); >> - ? ? } while (read_seqretry(&state->seqlock, seq)); >> - ? ? set_bit(NFS_LAYOUT_STATEID_SET, &lo->state); >> - ? ? write_sequnlock(&lo->seqlock); >> - ? ? dprintk("<-- %s\n", __func__); >> ?} >> >> ?/* Layoutreturn may use an invalid stateid, just copy what is there */ >> ?void pnfs_copy_layout_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo) >> ?{ >> - ? ? int seq; >> - >> - ? ? do { >> - ? ? ? ? ? ? seq = read_seqbegin(&lo->seqlock); >> - ? ? ? ? ? ? memcpy(dst->data, lo->stateid.data, sizeof(lo->stateid.data)); >> - ? ? } while (read_seqretry(&lo->seqlock, seq)); >> + ? ? assert_spin_locked(&lo->inode->i_lock); >> + ? ? memcpy(dst->data, lo->stateid.data, sizeof(lo->stateid.data)); > > This function is just redundant now. > Let's just open code its two users. > > Benny > OK. Fred >> ?} >> >> ?void >> ?pnfs_get_layout_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo, >> ? ? ? ? ? ? ? ? ? ? ? struct nfs4_state *open_state) >> ?{ >> - ? ? int seq; >> - >> ? ? ? dprintk("--> %s\n", __func__); >> - ? ? do { >> - ? ? ? ? ? ? seq = read_seqbegin(&lo->seqlock); >> - ? ? ? ? ? ? if (!test_bit(NFS_LAYOUT_STATEID_SET, &lo->state)) { >> - ? ? ? ? ? ? ? ? ? ? /* This will trigger retry of the read */ >> - ? ? ? ? ? ? ? ? ? ? pnfs_layout_from_open_stateid(lo, open_state); >> - ? ? ? ? ? ? } else >> - ? ? ? ? ? ? ? ? ? ? memcpy(dst->data, lo->stateid.data, >> - ? ? ? ? ? ? ? ? ? ? ? ? ? ?sizeof(lo->stateid.data)); >> - ? ? } while (read_seqretry(&lo->seqlock, seq)); >> + ? ? spin_lock(&lo->inode->i_lock); >> + ? ? if (!test_bit(NFS_LAYOUT_STATEID_SET, &lo->state)) { >> + ? ? ? ? ? ? int seq; >> + >> + ? ? ? ? ? ? do { >> + ? ? ? ? ? ? ? ? ? ? seq = read_seqbegin(&open_state->seqlock); >> + ? ? ? ? ? ? ? ? ? ? memcpy(dst->data, open_state->stateid.data, >> + ? ? ? ? ? ? ? ? ? ? ? ? ? ?sizeof(open_state->stateid.data)); >> + ? ? ? ? ? ? } while (read_seqretry(&open_state->seqlock, seq)); >> + ? ? ? ? ? ? set_bit(NFS_LAYOUT_STATEID_SET, &lo->state); >> + ? ? } else >> + ? ? ? ? ? ? memcpy(dst->data, lo->stateid.data, >> + ? ? ? ? ? ? ? ? ? ?sizeof(lo->stateid.data)); >> + ? ? spin_unlock(&lo->inode->i_lock); >> ? ? ? dprintk("<-- %s\n", __func__); >> ?} >> >> @@ -791,7 +771,6 @@ alloc_init_layout_hdr(struct inode *ino) >> ? ? ? lo->refcount = 1; >> ? ? ? INIT_LIST_HEAD(&lo->layouts); >> ? ? ? INIT_LIST_HEAD(&lo->segs); >> - ? ? seqlock_init(&lo->seqlock); >> ? ? ? lo->inode = ino; >> ? ? ? return lo; >> ?} >> diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h >> index 000acf0..de4eaa8 100644 >> --- a/fs/nfs/pnfs.h >> +++ b/fs/nfs/pnfs.h >> @@ -96,7 +96,6 @@ struct pnfs_layout_hdr { >> ? ? ? struct list_head ? ? ? ?layouts; ? /* other client layouts */ >> ? ? ? struct list_head ? ? ? ?segs; ? ? ?/* layout segments list */ >> ? ? ? int ? ? ? ? ? ? ? ? ? ? roc_iomode;/* return on close iomode, 0=none */ >> - ? ? seqlock_t ? ? ? ? ? ? ? seqlock; ? /* Protects the stateid */ >> ? ? ? nfs4_stateid ? ? ? ? ? ?stateid; >> ? ? ? unsigned long ? ? ? ? ? state; >> ? ? ? struct rpc_cred ? ? ? ? *cred; ? ? /* layoutcommit credential */ >> @@ -224,9 +223,8 @@ static inline int lo_fail_bit(u32 iomode) >> >> ?static inline void pnfs_invalidate_layout_stateid(struct pnfs_layout_hdr *lo) >> ?{ >> - ? ? write_seqlock(&lo->seqlock); >> + ? ? assert_spin_locked(&lo->inode->i_lock); >> ? ? ? clear_bit(NFS_LAYOUT_STATEID_SET, &lo->state); >> - ? ? write_sequnlock(&lo->seqlock); >> ?} >> >> ?static inline void get_lseg(struct pnfs_layout_segment *lseg) > -- > To unsubscribe from this list: send the line "unsubscribe linux-nfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at ?http://vger.kernel.org/majordomo-info.html >