Commit a6fb074a authored by Tomasz Zawadzki's avatar Tomasz Zawadzki
Browse files

lib/blob: avoid error paths after grow persists size 



Any errors before writing out super block are fine.
Meanwhile errors after, would require writing to
super block again with prior size. If not possible again,
handling it would become complex.
Avoid the error rather than trying to change
the super block/metadata format or adding second
persistent step.

This patch creates a bit pool and mask for used_clusters,
so that it can be used once super block is persisted.

Signed-off-by: default avatarTomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I554cb4ba078a5e97584afdcc17e9a40b4bdcd672
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/19793


Reviewed-by: default avatarJim Harris <jim.harris@samsung.com>
Tested-by: default avatarSPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: default avatarShuhei Matsumoto <smatsumoto@nvidia.com>
parent ffb60487

lib/blob/blobstore.c

+29 −10
Original line number Diff line number Diff line
@@ -8759,13 +8759,21 @@ struct spdk_bs_grow_ctx { 
	struct spdk_blob_store		*bs;

	struct spdk_bs_super_block	*super;



	struct spdk_bit_pool		*new_used_clusters;

	struct spdk_bs_md_mask		*new_used_clusters_mask;



	spdk_bs_sequence_t		*seq;

};



static void

bs_grow_live_done(struct spdk_bs_grow_ctx *ctx, int bserrno)

{

	if (bserrno != 0) {

		spdk_bit_pool_free(&ctx->new_used_clusters);

	}



	bs_sequence_finish(ctx->seq, bserrno);

	free(ctx->new_used_clusters_mask);

	spdk_free(ctx->super);

	free(ctx);

}
@@ -8776,7 +8784,6 @@ bs_grow_live_super_write_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno) 
	struct spdk_bs_grow_ctx	*ctx = cb_arg;

	struct spdk_blob_store *bs = ctx->bs;

	uint64_t total_clusters;

	int rc;



	if (bserrno != 0) {

		bs_grow_live_done(ctx, bserrno);
@@ -8793,19 +8800,20 @@ bs_grow_live_super_write_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno) 
	 */

	bs->clean = 0;



	/* Reverting the super->size past this point is complex, avoid any error paths

	 * that require to do so. */

	spdk_spin_lock(&bs->used_lock);



	total_clusters = ctx->super->size / ctx->super->cluster_size;



	rc = spdk_bit_pool_resize(&bs->used_clusters, total_clusters);

	if (rc < 0) {

		/* When this fails, super block contains new size that will only

		 * be accesible after blobstore reload. Error reported up is correct now,

		 * but does not apply after reload. */

		spdk_spin_unlock(&bs->used_lock);

		bs_grow_live_done(ctx, rc);

		return;

	}

	assert(total_clusters >= spdk_bit_pool_capacity(bs->used_clusters));

	spdk_bit_pool_store_mask(bs->used_clusters, ctx->new_used_clusters_mask);



	assert(total_clusters == spdk_bit_pool_capacity(ctx->new_used_clusters));

	spdk_bit_pool_load_mask(ctx->new_used_clusters, ctx->new_used_clusters_mask);



	spdk_bit_pool_free(&bs->used_clusters);

	bs->used_clusters = ctx->new_used_clusters;



	bs->total_clusters = total_clusters;

	bs->total_data_clusters = bs->total_clusters - spdk_divide_round_up(
@@ -8863,6 +8871,17 @@ bs_grow_live_load_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno) 


	SPDK_DEBUGLOG(blob, "Resizing blobstore\n");



	ctx->new_used_clusters_mask = calloc(1, total_clusters);

	if (!ctx->new_used_clusters_mask) {

		bs_grow_live_done(ctx, -ENOMEM);

		return;

	}

	ctx->new_used_clusters = spdk_bit_pool_create(total_clusters);

	if (!ctx->new_used_clusters) {

		bs_grow_live_done(ctx, -ENOMEM);

		return;

	}



	ctx->super->clean = 0;

	ctx->super->size = dev_size;

	ctx->super->used_cluster_mask_len = used_cluster_mask_len;