Commit df8e96f3 authored by David Woodhouse's avatar David Woodhouse
Browse files

[JFFS2] Improve read_inode memory usage, v2. 



We originally used to read every node and allocate a jffs2_tmp_dnode_info
structure for each, before processing them in (reverse) version order
and discarding the ones which are obsoleted by later nodes.

With huge logfiles, this behaviour caused memory problems. For example, a
file involved in OLPC trac #1292 has 1822391 nodes, and would cause the XO
machine to run out of memory during the first stage of read_inode().

Instead of just inserting nodes into a tree in version order as we find
them, we now put them into a tree in order of their offset within the
file, which allows us to immediately discard nodes which are completely
obsoleted.

We don't use a full tree with 'fragments' pointing to the real data
structure, as we do in the normal fragtree. We sort only on the start
address, and add an 'overlapped' flag to the tmp_dnode_info to indicate
that the node in question is (partially) overlapped by another.

When the scan is complete, we start at the end of the file, adding each
node to a real fragtree as before. Where the node is non-overlapped, we
just add it (it doesn't matter that it's not the latest version; there is
no overlap). When the node at the end of the tree _is_ overlapped, we sort
it and all its overlapping nodes into version order and then add them to
the fragtree in that order.

This 'early discard' reduces the peak allocation of tmp_dnode_info
structures from 1.8M to a mere 62872 (3.5%) in the degenerate case
referenced above.

This version of the patch also correctly rememembers the highest node
version# seen for an inode when it's scanned.

Signed-off-by: default avatarDavid Woodhouse <dwmw2@infradead.org>
parent 44b998e1

fs/jffs2/nodelist.c

+0 −460
Original line number Diff line number Diff line
@@ -397,466 +397,6 @@ int jffs2_add_full_dnode_to_inode(struct jffs2_sb_info *c, struct jffs2_inode_in 
	return 0;

}



/*

 * Check the data CRC of the node.

 *

 * Returns: 0 if the data CRC is correct;

 * 	    1 - if incorrect;

 *	    error code if an error occured.

 */

static int check_node_data(struct jffs2_sb_info *c, struct jffs2_tmp_dnode_info *tn)

{

	struct jffs2_raw_node_ref *ref = tn->fn->raw;

	int err = 0, pointed = 0;

	struct jffs2_eraseblock *jeb;

	unsigned char *buffer;

	uint32_t crc, ofs, len;

	size_t retlen;



	BUG_ON(tn->csize == 0);



	if (!jffs2_is_writebuffered(c))

		goto adj_acc;



	/* Calculate how many bytes were already checked */

	ofs = ref_offset(ref) + sizeof(struct jffs2_raw_inode);

	len = ofs % c->wbuf_pagesize;

	if (likely(len))

		len = c->wbuf_pagesize - len;



	if (len >= tn->csize) {

		dbg_readinode("no need to check node at %#08x, data length %u, data starts at %#08x - it has already been checked.\n",

			ref_offset(ref), tn->csize, ofs);

		goto adj_acc;

	}



	ofs += len;

	len = tn->csize - len;



	dbg_readinode("check node at %#08x, data length %u, partial CRC %#08x, correct CRC %#08x, data starts at %#08x, start checking from %#08x - %u bytes.\n",

		ref_offset(ref), tn->csize, tn->partial_crc, tn->data_crc, ofs - len, ofs, len);



#ifndef __ECOS

	/* TODO: instead, incapsulate point() stuff to jffs2_flash_read(),

	 * adding and jffs2_flash_read_end() interface. */

	if (c->mtd->point) {

		err = c->mtd->point(c->mtd, ofs, len, &retlen, &buffer);

		if (!err && retlen < tn->csize) {

			JFFS2_WARNING("MTD point returned len too short: %zu instead of %u.\n", retlen, tn->csize);

			c->mtd->unpoint(c->mtd, buffer, ofs, len);

		} else if (err)

			JFFS2_WARNING("MTD point failed: error code %d.\n", err);

		else

			pointed = 1; /* succefully pointed to device */

	}

#endif



	if (!pointed) {

		buffer = kmalloc(len, GFP_KERNEL);

		if (unlikely(!buffer))

			return -ENOMEM;



		/* TODO: this is very frequent pattern, make it a separate

		 * routine */

		err = jffs2_flash_read(c, ofs, len, &retlen, buffer);

		if (err) {

			JFFS2_ERROR("can not read %d bytes from 0x%08x, error code: %d.\n", len, ofs, err);

			goto free_out;

		}



		if (retlen != len) {

			JFFS2_ERROR("short read at %#08x: %zd instead of %d.\n", ofs, retlen, len);

			err = -EIO;

			goto free_out;

		}

	}



	/* Continue calculating CRC */

	crc = crc32(tn->partial_crc, buffer, len);

	if(!pointed)

		kfree(buffer);

#ifndef __ECOS

	else

		c->mtd->unpoint(c->mtd, buffer, ofs, len);

#endif



	if (crc != tn->data_crc) {

		JFFS2_NOTICE("wrong data CRC in data node at 0x%08x: read %#08x, calculated %#08x.\n",

			ofs, tn->data_crc, crc);

		return 1;

	}



adj_acc:

	jeb = &c->blocks[ref->flash_offset / c->sector_size];

	len = ref_totlen(c, jeb, ref);



	/*

	 * Mark the node as having been checked and fix the

	 * accounting accordingly.

	 */

	spin_lock(&c->erase_completion_lock);

	jeb->used_size += len;

	jeb->unchecked_size -= len;

	c->used_size += len;

	c->unchecked_size -= len;

	spin_unlock(&c->erase_completion_lock);



	return 0;



free_out:

	if(!pointed)

		kfree(buffer);

#ifndef __ECOS

	else

		c->mtd->unpoint(c->mtd, buffer, ofs, len);

#endif

	return err;

}



/*

 * Helper function for jffs2_add_older_frag_to_fragtree().

 *

 * Checks the node if we are in the checking stage.

 */

static int check_node(struct jffs2_sb_info *c, struct jffs2_inode_info *f, struct jffs2_tmp_dnode_info *tn)

{

	int ret;



	BUG_ON(ref_obsolete(tn->fn->raw));



	/* We only check the data CRC of unchecked nodes */

	if (ref_flags(tn->fn->raw) != REF_UNCHECKED)

		return 0;



	dbg_fragtree2("check node %#04x-%#04x, phys offs %#08x.\n",

		tn->fn->ofs, tn->fn->ofs + tn->fn->size, ref_offset(tn->fn->raw));



	ret = check_node_data(c, tn);

	if (unlikely(ret < 0)) {

		JFFS2_ERROR("check_node_data() returned error: %d.\n",

			ret);

	} else if (unlikely(ret > 0)) {

		dbg_fragtree2("CRC error, mark it obsolete.\n");

		jffs2_mark_node_obsolete(c, tn->fn->raw);

	}



	return ret;

}



/*

 * Helper function for jffs2_add_older_frag_to_fragtree().

 *

 * Called when the new fragment that is being inserted

 * splits a hole fragment.

 */

static int split_hole(struct jffs2_sb_info *c, struct rb_root *root,

		      struct jffs2_node_frag *newfrag, struct jffs2_node_frag *hole)

{

	dbg_fragtree2("fragment %#04x-%#04x splits the hole %#04x-%#04x\n",

		newfrag->ofs, newfrag->ofs + newfrag->size, hole->ofs, hole->ofs + hole->size);



	if (hole->ofs == newfrag->ofs) {

		/*

		 * Well, the new fragment actually starts at the same offset as

		 * the hole.

		 */

		if (hole->ofs + hole->size > newfrag->ofs + newfrag->size) {

			/*

			 * We replace the overlapped left part of the hole by

			 * the new node.

			 */



			dbg_fragtree2("insert fragment %#04x-%#04x and cut the left part of the hole\n",

				newfrag->ofs, newfrag->ofs + newfrag->size);

			rb_replace_node(&hole->rb, &newfrag->rb, root);



			hole->ofs += newfrag->size;

			hole->size -= newfrag->size;



			/*

			 * We know that 'hole' should be the right hand

			 * fragment.

			 */

			jffs2_fragtree_insert(hole, newfrag);

			rb_insert_color(&hole->rb, root);

		} else {

			/*

			 * Ah, the new fragment is of the same size as the hole.

			 * Relace the hole by it.

			 */

			dbg_fragtree2("insert fragment %#04x-%#04x and overwrite hole\n",

				newfrag->ofs, newfrag->ofs + newfrag->size);

			rb_replace_node(&hole->rb, &newfrag->rb, root);

			jffs2_free_node_frag(hole);

		}

	} else {

		/* The new fragment lefts some hole space at the left */



		struct jffs2_node_frag * newfrag2 = NULL;



		if (hole->ofs + hole->size > newfrag->ofs + newfrag->size) {

			/* The new frag also lefts some space at the right */

			newfrag2 = new_fragment(NULL, newfrag->ofs +

				newfrag->size, hole->ofs + hole->size

				- newfrag->ofs - newfrag->size);

			if (unlikely(!newfrag2)) {

				jffs2_free_node_frag(newfrag);

				return -ENOMEM;

			}

		}



		hole->size = newfrag->ofs - hole->ofs;

		dbg_fragtree2("left the hole %#04x-%#04x at the left and inserd fragment %#04x-%#04x\n",

			hole->ofs, hole->ofs + hole->size, newfrag->ofs, newfrag->ofs + newfrag->size);



		jffs2_fragtree_insert(newfrag, hole);

		rb_insert_color(&newfrag->rb, root);



		if (newfrag2) {

			dbg_fragtree2("left the hole %#04x-%#04x at the right\n",

				newfrag2->ofs, newfrag2->ofs + newfrag2->size);

			jffs2_fragtree_insert(newfrag2, newfrag);

			rb_insert_color(&newfrag2->rb, root);

		}

	}



	return 0;

}



/*

 * This function is used when we build inode. It expects the nodes are passed

 * in the decreasing version order. The whole point of this is to improve the

 * inodes checking on NAND: we check the nodes' data CRC only when they are not

 * obsoleted. Previously, add_frag_to_fragtree() function was used and

 * nodes were passed to it in the increasing version ordes and CRCs of all

 * nodes were checked.

 *

 * Note: tn->fn->size shouldn't be zero.

 *

 * Returns 0 if the node was inserted

 *         1 if it wasn't inserted (since it is obsolete)

 *         < 0 an if error occured

 */

int jffs2_add_older_frag_to_fragtree(struct jffs2_sb_info *c, struct jffs2_inode_info *f,

				     struct jffs2_tmp_dnode_info *tn)

{

	struct jffs2_node_frag *this, *newfrag;

	uint32_t lastend;

	struct jffs2_full_dnode *fn = tn->fn;

	struct rb_root *root = &f->fragtree;

	uint32_t fn_size = fn->size, fn_ofs = fn->ofs;

	int err, checked = 0;

	int ref_flag;



	dbg_fragtree("insert fragment %#04x-%#04x, ver %u\n", fn_ofs, fn_ofs + fn_size, tn->version);



	/* Skip all the nodes which are completed before this one starts */

	this = jffs2_lookup_node_frag(root, fn_ofs);

	if (this)

		dbg_fragtree2("'this' found %#04x-%#04x (%s)\n", this->ofs, this->ofs + this->size, this->node ? "data" : "hole");



	if (this)

		lastend = this->ofs + this->size;

	else

		lastend = 0;



	/* Detect the preliminary type of node */

	if (fn->size >= PAGE_CACHE_SIZE)

		ref_flag = REF_PRISTINE;

	else

		ref_flag = REF_NORMAL;



	/* See if we ran off the end of the root */

	if (lastend <= fn_ofs) {

		/* We did */



		/*

		 * We are going to insert the new node into the

		 * fragment tree, so check it.

		 */

		err = check_node(c, f, tn);

		if (err != 0)

			return err;



		fn->frags = 1;



		newfrag = new_fragment(fn, fn_ofs, fn_size);

		if (unlikely(!newfrag))

			return -ENOMEM;



		err = no_overlapping_node(c, root, newfrag, this, lastend);

		if (unlikely(err != 0)) {

			jffs2_free_node_frag(newfrag);

			return err;

		}



		goto out_ok;

	}



	fn->frags = 0;



	while (1) {

		/*

		 * Here we have:

		 * fn_ofs < this->ofs + this->size && fn_ofs >= this->ofs.

		 *

		 * Remember, 'this' has higher version, any non-hole node

		 * which is already in the fragtree is newer then the newly

		 * inserted.

		 */

		if (!this->node) {

			/*

			 * 'this' is the hole fragment, so at least the

			 * beginning of the new fragment is valid.

			 */



			/*

			 * We are going to insert the new node into the

			 * fragment tree, so check it.

			 */

			if (!checked) {

				err = check_node(c, f, tn);

				if (unlikely(err != 0))

					return err;

				checked = 1;

			}



			if (this->ofs + this->size >= fn_ofs + fn_size) {

				/* We split the hole on two parts */



				fn->frags += 1;

				newfrag = new_fragment(fn, fn_ofs, fn_size);

				if (unlikely(!newfrag))

					return -ENOMEM;



				err = split_hole(c, root, newfrag, this);

				if (unlikely(err))

					return err;

				goto out_ok;

			}



			/*

			 * The beginning of the new fragment is valid since it

			 * overlaps the hole node.

			 */



			ref_flag = REF_NORMAL;



			fn->frags += 1;

			newfrag = new_fragment(fn, fn_ofs,

					this->ofs + this->size - fn_ofs);

			if (unlikely(!newfrag))

				return -ENOMEM;



			if (fn_ofs == this->ofs) {

				/*

				 * The new node starts at the same offset as

				 * the hole and supersieds the hole.

				 */

				dbg_fragtree2("add the new fragment instead of hole %#04x-%#04x, refcnt %d\n",

					fn_ofs, fn_ofs + this->ofs + this->size - fn_ofs, fn->frags);



				rb_replace_node(&this->rb, &newfrag->rb, root);

				jffs2_free_node_frag(this);

			} else {

				/*

				 * The hole becomes shorter as its right part

				 * is supersieded by the new fragment.

				 */

				dbg_fragtree2("reduce size of hole %#04x-%#04x to %#04x-%#04x\n",

					this->ofs, this->ofs + this->size, this->ofs, this->ofs + this->size - newfrag->size);



				dbg_fragtree2("add new fragment %#04x-%#04x, refcnt %d\n", fn_ofs,

					fn_ofs + this->ofs + this->size - fn_ofs, fn->frags);



				this->size -= newfrag->size;

				jffs2_fragtree_insert(newfrag, this);

				rb_insert_color(&newfrag->rb, root);

			}



			fn_ofs += newfrag->size;

			fn_size -= newfrag->size;

			this = rb_entry(rb_next(&newfrag->rb),

					struct jffs2_node_frag, rb);



			dbg_fragtree2("switch to the next 'this' fragment: %#04x-%#04x %s\n",

				this->ofs, this->ofs + this->size, this->node ? "(data)" : "(hole)");

		}



		/*

		 * 'This' node is not the hole so it obsoletes the new fragment

		 * either fully or partially.

		 */

		if (this->ofs + this->size >= fn_ofs + fn_size) {

			/* The new node is obsolete, drop it */

			if (fn->frags == 0) {

				dbg_fragtree2("%#04x-%#04x is obsolete, mark it obsolete\n", fn_ofs, fn_ofs + fn_size);

				ref_flag = REF_OBSOLETE;

			}

			goto out_ok;

		} else {

			struct jffs2_node_frag *new_this;



			/* 'This' node obsoletes the beginning of the new node */

			dbg_fragtree2("the beginning %#04x-%#04x is obsolete\n", fn_ofs, this->ofs + this->size);



			ref_flag = REF_NORMAL;



			fn_size -= this->ofs + this->size - fn_ofs;

			fn_ofs = this->ofs + this->size;

			dbg_fragtree2("now considering %#04x-%#04x\n", fn_ofs, fn_ofs + fn_size);



			new_this = rb_entry(rb_next(&this->rb), struct jffs2_node_frag, rb);

			if (!new_this) {

				/*

				 * There is no next fragment. Add the rest of

				 * the new node as the right-hand child.

				 */

				if (!checked) {

					err = check_node(c, f, tn);

					if (unlikely(err != 0))

						return err;

					checked = 1;

				}



				fn->frags += 1;

				newfrag = new_fragment(fn, fn_ofs, fn_size);

				if (unlikely(!newfrag))

					return -ENOMEM;



				dbg_fragtree2("there are no more fragments, insert %#04x-%#04x\n",

					newfrag->ofs, newfrag->ofs + newfrag->size);

				rb_link_node(&newfrag->rb, &this->rb, &this->rb.rb_right);

				rb_insert_color(&newfrag->rb, root);

				goto out_ok;

			} else {

				this = new_this;

				dbg_fragtree2("switch to the next 'this' fragment: %#04x-%#04x %s\n",

					this->ofs, this->ofs + this->size, this->node ? "(data)" : "(hole)");

			}

		}

	}



out_ok:

	BUG_ON(fn->size < PAGE_CACHE_SIZE && ref_flag == REF_PRISTINE);



	if (ref_flag == REF_OBSOLETE) {

		dbg_fragtree2("the node is obsolete now\n");

		/* jffs2_mark_node_obsolete() will adjust space accounting */

		jffs2_mark_node_obsolete(c, fn->raw);

		return 1;

	}



	dbg_fragtree2("the node is \"%s\" now\n", ref_flag == REF_NORMAL ? "REF_NORMAL" : "REF_PRISTINE");



	/* Space accounting was adjusted at check_node_data() */

	spin_lock(&c->erase_completion_lock);

	fn->raw->flash_offset = ref_offset(fn->raw) | ref_flag;

	spin_unlock(&c->erase_completion_lock);



	return 0;

}



void jffs2_set_inocache_state(struct jffs2_sb_info *c, struct jffs2_inode_cache *ic, int state)

{

	spin_lock(&c->inocache_lock);

fs/jffs2/nodelist.h

+23 −2
Original line number Diff line number Diff line
@@ -225,7 +225,20 @@ struct jffs2_tmp_dnode_info 
	uint32_t version;

	uint32_t data_crc;

	uint32_t partial_crc;

	uint32_t csize;

	uint16_t csize;

	uint16_t overlapped;

};



/* Temporary data structure used during readinode. */

struct jffs2_readinode_info

{

	struct rb_root tn_root;

	struct jffs2_tmp_dnode_info *mdata_tn;

	uint32_t highest_version;

	uint32_t latest_mctime;

	uint32_t mctime_ver;

	struct jffs2_full_dirent *fds;

	struct jffs2_raw_node_ref *latest_ref;

};



struct jffs2_full_dirent
@@ -328,6 +341,15 @@ static inline struct jffs2_node_frag *frag_last(struct rb_root *root) 
#define frag_right(frag) rb_entry((frag)->rb.rb_right, struct jffs2_node_frag, rb)

#define frag_erase(frag, list) rb_erase(&frag->rb, list);



#define tn_next(tn) rb_entry(rb_next(&(tn)->rb), struct jffs2_tmp_dnode_info, rb)

#define tn_prev(tn) rb_entry(rb_prev(&(tn)->rb), struct jffs2_tmp_dnode_info, rb)

#define tn_parent(tn) rb_entry(rb_parent(&(tn)->rb), struct jffs2_tmp_dnode_info, rb)

#define tn_left(tn) rb_entry((tn)->rb.rb_left, struct jffs2_tmp_dnode_info, rb)

#define tn_right(tn) rb_entry((tn)->rb.rb_right, struct jffs2_tmp_dnode_info, rb)

#define tn_erase(tn, list) rb_erase(&tn->rb, list);

#define tn_last(list) rb_entry(rb_last(list), struct jffs2_tmp_dnode_info, rb)

#define tn_first(list) rb_entry(rb_first(list), struct jffs2_tmp_dnode_info, rb)



/* nodelist.c */

void jffs2_add_fd_to_list(struct jffs2_sb_info *c, struct jffs2_full_dirent *new, struct jffs2_full_dirent **list);

void jffs2_set_inocache_state(struct jffs2_sb_info *c, struct jffs2_inode_cache *ic, int state);
@@ -343,7 +365,6 @@ struct rb_node *rb_prev(struct rb_node *); 
void rb_replace_node(struct rb_node *victim, struct rb_node *new, struct rb_root *root);

int jffs2_add_full_dnode_to_inode(struct jffs2_sb_info *c, struct jffs2_inode_info *f, struct jffs2_full_dnode *fn);

void jffs2_truncate_fragtree (struct jffs2_sb_info *c, struct rb_root *list, uint32_t size);

int jffs2_add_older_frag_to_fragtree(struct jffs2_sb_info *c, struct jffs2_inode_info *f, struct jffs2_tmp_dnode_info *tn);

struct jffs2_raw_node_ref *jffs2_link_node_ref(struct jffs2_sb_info *c,

					       struct jffs2_eraseblock *jeb,

					       uint32_t ofs, uint32_t len,

fs/jffs2/readinode.c

+595 −137

File changed.

Preview size limit exceeded, changes collapsed.

CRA Git | Maintained and supported by SUSTech CRA and CCSE