LU-4931 ladvise: Add dontneed advice support for ladvise

This patch addds DONTNEED advice to ladvise framework. OSS will
cleanup the page cache of the file when this hint is provided.

Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Gu Zheng <gzheng@ddn.com>
Change-Id: If5cf7f3193924ca7cccb96d8d841c0d889469393
Reviewed-on: http://review.whamcloud.com/20203
Reviewed-by: Wang Shilong <wshilong@ddn.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/doc/lfs-ladvise.1 b/lustre/doc/lfs-ladvise.1
index 295658b..6810a14 100644 (file)
--- a/lustre/doc/lfs-ladvise.1
+++ b/lustre/doc/lfs-ladvise.1
@@ -19,6 +19,8 @@ Give advice or hint of type \fIADVICE\fR. Advice types are:
 .RS 1.2i
 .TP
 \fBwillread\fR to prefetch data into server cache
+.TP
+\fBdontneed\fR to cleanup data cache on server
 .RE
 .TP
 \fB\-b\fR, \fB\-\-background
@@ -58,6 +60,11 @@ filesystem, while ladvise can send advices or hints to Lustre server sides.
 .B $ lfs ladvise -a willread -s 0 -e 1048576000 /mnt/lustre/file1
 This gives the OST(s) holding the first 1GB of \fB/mnt/lustre/file1\fR a hint
 that the first 1GB of the file will be read soon.
+.TP
+.B $ lfs ladvise -a dontneed -s 0 -e 1048576000 /mnt/lustre/file1
+This gives the OST(s) holding the first 1GB of \fB/mnt/lustre/file1\fR a hint
+that the first 1GB of file will not be read in the near future, thus the OST(s)
+could clear the cache of the file in the memory.
 .SH AVAILABILITY
 The lfs ladvise command is part of the Lustre filesystem.
 .SH SEE ALSO

diff --git a/lustre/include/lustre/lustre_user.h b/lustre/include/lustre/lustre_user.h
index 6fe9ae4..a27b876 100644 (file)
--- a/lustre/include/lustre/lustre_user.h
+++ b/lustre/include/lustre/lustre_user.h
@@ -1379,10 +1379,12 @@ struct llapi_json_item_list {
 enum lu_ladvise_type {
        LU_LADVISE_INVALID      = 0,
        LU_LADVISE_WILLREAD     = 1,
+       LU_LADVISE_DONTNEED     = 2,
 };
 
 #define LU_LADVISE_NAMES {                                             \
        [LU_LADVISE_WILLREAD]   = "willread",                           \
+       [LU_LADVISE_DONTNEED]   = "dontneed",                           \
 }
 
 /* This is the userspace argument for ladvise.  It is currently the same as

diff --git a/lustre/ofd/ofd_dev.c b/lustre/ofd/ofd_dev.c
index cfa2baa..40cbf71 100644 (file)
--- a/lustre/ofd/ofd_dev.c
+++ b/lustre/ofd/ofd_dev.c
@@ -2191,6 +2191,7 @@ static int ofd_ladvise_hdl(struct tgt_session_info *tsi)
        struct lustre_handle     lockh = { 0 };
        __u64                    flags = 0;
        int                      i;
+       struct dt_object        *dob;
        ENTRY;
 
        CFS_FAIL_TIMEOUT(OBD_FAIL_OST_LADVISE_PAUSE, cfs_fail_val);
@@ -2237,6 +2238,7 @@ static int ofd_ladvise_hdl(struct tgt_session_info *tsi)
                RETURN(rc);
        }
        LASSERT(fo != NULL);
+       dob = ofd_object_child(fo);
 
        for (i = 0; i < num_advise; i++, ladvise++) {
                if (ladvise->lla_end <= ladvise->lla_start) {
@@ -2266,6 +2268,10 @@ static int ofd_ladvise_hdl(struct tgt_session_info *tsi)
                                ladvise->lla_end);
                        tgt_extent_unlock(&lockh, LCK_PR);
                        break;
+               case LU_LADVISE_DONTNEED:
+                       rc = dt_ladvise(env, dob, ladvise->lla_start,
+                                       ladvise->lla_end, LU_LADVISE_DONTNEED);
+                       break;
                }
                if (rc != 0)
                        break;

diff --git a/lustre/osd-ldiskfs/osd_io.c b/lustre/osd-ldiskfs/osd_io.c
index 91034d2..9bfe5aa 100644 (file)
--- a/lustre/osd-ldiskfs/osd_io.c
+++ b/lustre/osd-ldiskfs/osd_io.c
@@ -1933,10 +1933,18 @@ static int osd_fiemap_get(const struct lu_env *env, struct dt_object *dt,
 static int osd_ladvise(const struct lu_env *env, struct dt_object *dt,
                       __u64 start, __u64 end, enum lu_ladvise_type advice)
 {
-       int     rc;
+       int              rc = 0;
+       struct inode    *inode = osd_dt_obj(dt)->oo_inode;
        ENTRY;
 
        switch (advice) {
+       case LU_LADVISE_DONTNEED:
+               if (end == 0)
+                       break;
+               invalidate_mapping_pages(inode->i_mapping,
+                                        start >> PAGE_CACHE_SHIFT,
+                                        (end - 1) >> PAGE_CACHE_SHIFT);
+               break;
        default:
                rc = -ENOTSUPP;
                break;

diff --git a/lustre/ptlrpc/wiretest.c b/lustre/ptlrpc/wiretest.c
index 6f383dd..5af7e02 100644 (file)
--- a/lustre/ptlrpc/wiretest.c
+++ b/lustre/ptlrpc/wiretest.c
@@ -5015,6 +5015,8 @@ void lustre_assert_wire_constants(void)
                 (long long)(int)sizeof(((struct lu_ladvise *)0)->lla_value4));
        LASSERTF(LU_LADVISE_WILLREAD == 1, "found %lld\n",
                 (long long)LU_LADVISE_WILLREAD);
+       LASSERTF(LU_LADVISE_DONTNEED == 2, "found %lld\n",
+                (long long)LU_LADVISE_DONTNEED);
 
        /* Checks for struct ladvise_hdr */
        LASSERTF(LADVISE_MAGIC == 0x1ADF1CE0, "found 0x%.8x\n",

diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh
index 507812d..0c617fe 100755 (executable)
--- a/lustre/tests/sanity.sh
+++ b/lustre/tests/sanity.sh
@@ -14226,6 +14226,71 @@ test_255a() {
 }
 run_test 255a "check 'lfs ladvise -a willread'"
 
+facet_meminfo() {
+       local facet=$1
+       local info=$2
+
+       do_facet $facet "cat /proc/meminfo | grep ^${info}:" | awk '{print $2}'
+}
+
+test_255b() {
+       lfs setstripe -c -1 -i 0 $DIR/$tfile
+
+       ladvise_no_type dontneed $DIR/$tfile &&
+               skip "dontneed ladvise is not supported" && return
+
+       ladvise_no_ioctl $DIR/$tfile &&
+               skip "ladvise ioctl is not supported" && return
+
+       [ $(lustre_version_code ost1) -lt $(version_code 2.8.54) ] &&
+               skip "lustre < 2.8.54 does not support ladvise" && return
+
+       [ "$(facet_fstype ost1)" = "zfs" ] &&
+               skip "zfs-osd does not support dontneed advice" && return
+
+       local size_mb=100
+       local size=$((size_mb * 1048576))
+       # In order to prevent disturbance of other processes, only check 3/4
+       # of the memory usage
+       local kibibytes=$((size_mb * 1024 * 3 / 4))
+
+       dd if=/dev/zero of=$DIR/$tfile bs=1048576 count=$size_mb ||
+               error "dd to $DIR/$tfile failed"
+
+       local total=$(facet_meminfo ost1 MemTotal)
+       echo "Total memory: $total KiB"
+
+       do_facet ost1 "sync && echo 3 > /proc/sys/vm/drop_caches"
+       local before_read=$(facet_meminfo ost1 Cached)
+       echo "Cache used before read: $before_read KiB"
+
+       lfs ladvise -a willread $DIR/$tfile ||
+               error "Ladvise willread failed"
+       local after_read=$(facet_meminfo ost1 Cached)
+       echo "Cache used after read: $after_read KiB"
+
+       lfs ladvise -a dontneed $DIR/$tfile ||
+               error "Ladvise dontneed again failed"
+       local no_read=$(facet_meminfo ost1 Cached)
+       echo "Cache used after dontneed ladvise: $no_read KiB"
+
+       if [ $total -lt $((before_read + kibibytes)) ]; then
+               echo "Memory is too small, abort checking"
+               return 0
+       fi
+
+       if [ $((before_read + kibibytes)) -gt $after_read ]; then
+               error "Ladvise willread should use more memory" \
+                       "than $kibibytes KiB"
+       fi
+
+       if [ $((no_read + kibibytes)) -gt $after_read ]; then
+               error "Ladvise dontneed should release more memory" \
+                       "than $kibibytes KiB"
+       fi
+}
+run_test 255b "check 'lfs ladvise -a dontneed'"
+
 test_256() {
        local cl_user
        local cat_sl

diff --git a/lustre/utils/wirecheck.c b/lustre/utils/wirecheck.c
index f2e63d6..fcaa88a 100644 (file)
--- a/lustre/utils/wirecheck.c
+++ b/lustre/utils/wirecheck.c
@@ -324,6 +324,7 @@ check_lu_ladvise(void)
        CHECK_MEMBER(lu_ladvise, lla_value3);
        CHECK_MEMBER(lu_ladvise, lla_value4);
        CHECK_VALUE(LU_LADVISE_WILLREAD);
+       CHECK_VALUE(LU_LADVISE_DONTNEED);
 
        CHECK_VALUE(LF_ASYNC);
        CHECK_VALUE(LADVISE_MAGIC);

diff --git a/lustre/utils/wiretest.c b/lustre/utils/wiretest.c
index 2dce67d..d28f60f 100644 (file)
--- a/lustre/utils/wiretest.c
+++ b/lustre/utils/wiretest.c
@@ -5030,6 +5030,8 @@ void lustre_assert_wire_constants(void)
                 (long long)(int)sizeof(((struct lu_ladvise *)0)->lla_value4));
        LASSERTF(LU_LADVISE_WILLREAD == 1, "found %lld\n",
                 (long long)LU_LADVISE_WILLREAD);
+       LASSERTF(LU_LADVISE_DONTNEED == 2, "found %lld\n",
+                (long long)LU_LADVISE_DONTNEED);
 
        /* Checks for struct ladvise_hdr */
        LASSERTF(LADVISE_MAGIC == 0x1ADF1CE0, "found 0x%.8x\n",