Merge commit 'refs/remotes/origin/linux-kernel-mem' into HEAD

.topdeps

@@ -1,3 +1 @@
-gcc-branch
+zfs-branch
-fix-branch
-feature-branch

.topmsg

@@ -1,19 +1,6 @@
 From: Brian Behlendorf <behlendorf1@llnl.gov>
-Subject: [PATCH] zfs branch
+Subject: [PATCH] linux kernel mem
 
-Merged result of all changes which are relevant to both Solaris
+Required kmem/vmem changes
-and Linux builds of the ZFS code.  These are changes where there
-is a reasonable chance they will be accepted upstream.
-
-Additionally, since this is effectively the root of the linux
-ZFS tree the core linux build system is added here.  This
-includes autogen.sh, configure.ac, m4 macros, some scripts/*,
-and makefiles for all the core ZFS components.  Linux-only
-features which require tweaks to the build system should appear
-on the relevant topic branches.  All autotools products which
-result from autogen.sh are commited to the linux-configure-branch.
-
-This branch also contains the META, ChangeLog, AUTHORS, TODO,
-and README, files.
 
 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

lib/libzpool/include/sys/zfs_context.h

@@ -303,6 +303,9 @@ extern void kstat_delete(kstat_t *);
 #define	kmem_alloc(_s, _f)	umem_alloc(_s, _f)
 #define	kmem_zalloc(_s, _f)	umem_zalloc(_s, _f)
 #define	kmem_free(_b, _s)	umem_free(_b, _s)
+#define	vmem_alloc(_s, _f)	kmem_alloc(_s, _f)
+#define	vmem_zalloc(_s, _f)	kmem_zalloc(_s, _f)
+#define	vmem_free(_b, _s)	kmem_free(_b, _s)
 #define	kmem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i) \
 	umem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i)
 #define	kmem_cache_destroy(_c)	umem_cache_destroy(_c)

module/zfs/arc.c

@@ -755,8 +755,15 @@ buf_fini(void)
 {
 	int i;
 
+#if defined(_KERNEL) && defined(HAVE_SPL)
+	/* Large allocations which do not require contiguous pages
+	 * should be using vmem_free() in the linux kernel */
+	vmem_free(buf_hash_table.ht_table,
+	    (buf_hash_table.ht_mask + 1) * sizeof (void *));
+#else
 	kmem_free(buf_hash_table.ht_table,
 	    (buf_hash_table.ht_mask + 1) * sizeof (void *));
+#endif
 	for (i = 0; i < BUF_LOCKS; i++)
 		mutex_destroy(&buf_hash_table.ht_locks[i].ht_lock);
 	kmem_cache_destroy(hdr_cache);
@@ -855,8 +862,15 @@ buf_init(void)
 		hsize <<= 1;
 retry:
 	buf_hash_table.ht_mask = hsize - 1;
+#if defined(_KERNEL) && defined(HAVE_SPL)
+	/* Large allocations which do not require contiguous pages
+	 * should be using vmem_alloc() in the linux kernel */
+	buf_hash_table.ht_table =
+	    vmem_zalloc(hsize * sizeof (void*), KM_SLEEP);
+#else
 	buf_hash_table.ht_table =
 	    kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP);
+#endif
 	if (buf_hash_table.ht_table == NULL) {
 		ASSERT(hsize > (1ULL << 8));
 		hsize >>= 1;

module/zfs/dbuf.c

@@ -255,7 +255,13 @@ dbuf_init(void)
 
 retry:
 	h->hash_table_mask = hsize - 1;
+#if defined(_KERNEL) && defined(HAVE_SPL)
+	/* Large allocations which do not require contiguous pages
+	 * should be using vmem_alloc() in the linux kernel */
+	h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP);
+#else
 	h->hash_table = kmem_zalloc(hsize * sizeof (void *), KM_NOSLEEP);
+#endif
 	if (h->hash_table == NULL) {
 		/* XXX - we should really return an error instead of assert */
 		ASSERT(hsize > (1ULL << 10));
@@ -279,7 +285,13 @@ dbuf_fini(void)
 
 	for (i = 0; i < DBUF_MUTEXES; i++)
 		mutex_destroy(&h->hash_mutexes[i]);
+#if defined(_KERNEL) && defined(HAVE_SPL)
+	/* Large allocations which do not require contiguous pages
+	 * should be using vmem_free() in the linux kernel */
+	vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
+#else
 	kmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
+#endif
 	kmem_cache_destroy(dbuf_cache);
 }
 

module/zfs/dmu_send.c

@@ -1008,7 +1008,7 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, vnode_t *vp, offset_t *voffp)
 	ra.vp = vp;
 	ra.voff = *voffp;
 	ra.bufsize = 1<<20;
-	ra.buf = kmem_alloc(ra.bufsize, KM_SLEEP);
+	ra.buf = vmem_alloc(ra.bufsize, KM_SLEEP);
 
 	/* these were verified in dmu_recv_begin */
 	ASSERT(drc->drc_drrb->drr_version == DMU_BACKUP_STREAM_VERSION);
@@ -1098,7 +1098,7 @@ out:
 		dmu_recv_abort_cleanup(drc);
 	}
 
-	kmem_free(ra.buf, ra.bufsize);
+	vmem_free(ra.buf, ra.bufsize);
 	*voffp = ra.voff;
 	return (ra.err);
 }

module/zfs/spa_history.c

@@ -281,7 +281,7 @@ spa_history_log_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
 	dmu_buf_rele(dbp, FTAG);
 
 	if (hap->ha_log_type == LOG_INTERNAL) {
-		kmem_free((void*)hap->ha_history_str, HIS_MAX_RECORD_LEN);
+		vmem_free((void*)hap->ha_history_str, HIS_MAX_RECORD_LEN);
 		kmem_free(hap, sizeof (history_arg_t));
 	}
 }
@@ -407,7 +407,7 @@ spa_history_internal_log(history_internal_events_t event, spa_t *spa,
 		return;
 
 	hap = kmem_alloc(sizeof (history_arg_t), KM_SLEEP);
-	str = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
+	str = vmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 
 	va_start(adx, fmt);
 	(void) vsnprintf(str, HIS_MAX_RECORD_LEN, fmt, adx);

module/zfs/txg.c

@@ -49,7 +49,7 @@ txg_init(dsl_pool_t *dp, uint64_t txg)
 	int c;
 	bzero(tx, sizeof (tx_state_t));
 
-	tx->tx_cpu = kmem_zalloc(max_ncpus * sizeof (tx_cpu_t), KM_SLEEP);
+	tx->tx_cpu = vmem_zalloc(max_ncpus * sizeof (tx_cpu_t), KM_SLEEP);
 
 	for (c = 0; c < max_ncpus; c++) {
 		int i;
@@ -109,7 +109,7 @@ txg_fini(dsl_pool_t *dp)
 	if (tx->tx_commit_cb_taskq != NULL)
 		taskq_destroy(tx->tx_commit_cb_taskq);
 
-	kmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
+	vmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
 
 	bzero(tx, sizeof (tx_state_t));
 }

module/zfs/zfs_ioctl.c

@@ -134,7 +134,7 @@ __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 static void
 history_str_free(char *buf)
 {
-	kmem_free(buf, HIS_MAX_RECORD_LEN);
+	vmem_free(buf, HIS_MAX_RECORD_LEN);
 }
 
 static char *
@@ -145,7 +145,7 @@ history_str_get(zfs_cmd_t *zc)
 	if (zc->zc_history == 0)
 		return (NULL);
 
-	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
+	buf = vmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
 		history_str_free(buf);
@@ -729,12 +729,12 @@ put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
 	if (size > zc->zc_nvlist_dst_size) {
 		error = ENOMEM;
 	} else {
-		packed = kmem_alloc(size, KM_SLEEP);
+		packed = vmem_alloc(size, KM_SLEEP);
 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
 		    KM_SLEEP) == 0);
 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
 		    size);
-		kmem_free(packed, size);
+		vmem_free(packed, size);
 	}
 
 	zc->zc_nvlist_dst_size = size;
@@ -3002,7 +3002,7 @@ zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
 		return (EINVAL);
 
-	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
+	zc = vmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
 
 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
 
@@ -3041,7 +3041,7 @@ zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
 			zfs_log_history(zc);
 	}
 
-	kmem_free(zc, sizeof (zfs_cmd_t));
+	vmem_free(zc, sizeof (zfs_cmd_t));
 	return (error);
 }