[#LU-6378] Quota performance issue for 2.7

/** * Helper function for osd_object_create() * * \retval 0, on success */ static int __osd_object_create(struct osd_thread_info *info, struct osd_object *obj, struct lu_attr *attr, struct dt_allocation_hint *hint, struct dt_object_format *dof, struct thandle *th) { int result; __u32 umask; /* we drop umask so that permissions we pass are not affected */ umask = current->fs->umask; current->fs->umask = 0; result = osd_create_type_f(dof->dof_type)(info, obj, attr, hint, dof, th); if (result == 0) { osd_attr_init(info, obj, attr, dof); osd_object_init0(obj); } if (obj->oo_inode != NULL) { LASSERT(obj->oo_inode->i_state & I_NEW); unlock_new_inode(obj->oo_inode); --->It will unlock new inode here, so it will cause the above osd_attr_init to call quota_initialization (with I_NEW). But in 2.6.0, we used to unlock the new inode in ldiskfs_new_inode. } /* restore previous umask value */ current->fs->umask = umask; return result; }

I think the correct fix is as Niu writes. The patch that was landed upstream didn't check I_NEW, but instead directly checked if the inode already had the quota initialized. The lustre/kernel_patches/patches/quota-avoid-dqget-calls patches should be updated to use the same code as was committed to the upstream kernel.

commit 1ea06bec78a128adc995ca32bd906a6c9bb9cf91
Author: Niu Yawei <yawei.niu@gmail.com>
Date:   Wed Jun 4 12:20:30 2014 +0800

    quota: avoid unnecessary dqget()/dqput() calls
    
    Avoid unnecessary dqget()/dqput() calls in __dquot_initialize(),
    that will introduce global lock contention otherwise.
    
    Signed-off-by: Lai Siyao <lai.siyao@intel.com>
    Signed-off-by: Niu Yawei <yawei.niu@intel.com>
    Signed-off-by: Jan Kara <jack@suse.cz>

diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c
index 7f30bdc..2517719 100644
--- a/fs/quota/dquot.c
+++ b/fs/quota/dquot.c
@@ -1402,7 +1402,7 @@ static int dquot_active(const struct inode *inode)
  */
 static void __dquot_initialize(struct inode *inode, int type)
 {
-       int cnt;
+       int cnt, init_needed = 0;
        struct dquot *got[MAXQUOTAS];
        struct super_block *sb = inode->i_sb;
        qsize_t rsv;
@@ -1418,6 +1418,15 @@ static void __dquot_initialize(struct inode *inode, int type)
                got[cnt] = NULL;
                if (type != -1 && cnt != type)
                        continue;
+               /*
+                * The i_dquot should have been initialized in most cases,
+                * we check it without locking here to avoid unnecessary
+                * dqget()/dqput() calls.
+                */
+               if (inode->i_dquot[cnt])
+                       continue;
+               init_needed = 1;
+
                switch (cnt) {
                case USRQUOTA:
                        qid = make_kqid_uid(inode->i_uid);
@@ -1429,6 +1438,10 @@ static void __dquot_initialize(struct inode *inode, int type)
                got[cnt] = dqget(sb, qid);
        }
 
+       /* All required i_dquot has been initialized */
+       if (!init_needed)
+               return;
+
        down_write(&sb_dqopt(sb)->dqptr_sem);
        if (IS_NOQUOTA(inode))
                goto out_err;

[LU-6378] Quota performance issue for 2.7 Created: 18/Mar/15 Updated: 12/May/16 Resolved: 29/Apr/15
Status:	Resolved
Project:	Lustre
Component/s:	None
Affects Version/s:	Lustre 2.7.0, Lustre 2.8.0
Fix Version/s:	Lustre 2.8.0

[LU-6378] Quota performance issue for 2.7 Created: 18/Mar/15 Updated: 12/May/16 Resolved: 29/Apr/15