[PATCHv2 9/9] common/attr: Reduce MAX_ATTRS to leave some overhead for 64K blocksize

From: Ritesh Harjani <riteshh@linux.ibm.com>
To: fstests@vger.kernel.org
Cc: linux-ext4@vger.kernel.org, "Theodore Ts'o" <tytso@mit.edu>,
	"Darrick J . Wong" <djwong@kernel.org>,
	Ritesh Harjani <riteshh@linux.ibm.com>
Subject: [PATCHv2 9/9] common/attr: Reduce MAX_ATTRS to leave some overhead for 64K blocksize
Date: Wed, 21 Jul 2021 10:58:02 +0530	[thread overview]
Message-ID: <1109c811f550f918b8ea8bbe49323f32653b488f.1626844259.git.riteshh@linux.ibm.com> (raw)
In-Reply-To: <cover.1626844259.git.riteshh@linux.ibm.com>

Test generic/020 fails for ext4 with 64K blocksize.
This adds changes in common/attr for MAX_ATTRS calculations for
ext2|ext3|ext4 along with comments explaining the calculations.

Suggested-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Ritesh Harjani <riteshh@linux.ibm.com>
---
 common/attr | 39 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/common/attr b/common/attr
index d3902346..35682d7c 100644
--- a/common/attr
+++ b/common/attr
@@ -256,6 +256,45 @@ case "$FSTYP" in
 xfs|udf|pvfs2|9p|ceph|nfs)
 	MAX_ATTRS=1000
 	;;
+ext2|ext3|ext4)
+	# For 4k blocksizes, most of the attributes have an attr_name of
+	# "attribute_NN" which is 12, and "value_NN" which is 8.
+	# But for larger block sizes, we start having extended attributes of the
+	# form "attribute_NNN" or "attribute_NNNN", and "value_NNN" and
+	# "value_NNNN", which causes the round(len(..), 4) to jump up by 4
+	# bytes.  So round_up(len(attr_name, 4)) becomes 16 instead of 12, and
+	# round_up(len(value, 4)) becomes 12 instead of 8.
+	#
+	# For 64K blocksize the calculation becomes
+	# 	max_attrs = (block_size - 32) / (16 + 12 + 16)
+	# or
+	# 	max_attrs = (block_size - 32) / 44
+	#
+	# For 4K blocksize:-
+	# 	max_attrs = (block_size - 32) / (16 + 8 + 12)
+	# or
+	# 	max_attrs = (block_size - 32) / 36
+	#
+	# Note (for 4K bs) above are exact calculations for attrs of type
+	# attribute_NN with values of type value_NN.
+	# With above calculations, for 4k blocksize max_attrs becomes 112.
+	# This means we can have few attrs of type attribute_NNN with values of
+	# type value_NNN. To avoid/handle this we need to add extra 4 bytes of
+	# headroom.
+	#
+	# So for 4K, the calculations becomes:-
+	# 	max_attrs = (block_size - 32) / (16 + 8 + 12 + 4)
+	# or
+	# 	max_attrs = (block_size - 32) / 40
+	#
+	# Assume max ~1 block of attrs
+	BLOCK_SIZE=`_get_block_size $TEST_DIR`
+	if [ $BLOCK_SIZE -le 4096 ]; then
+		let MAX_ATTRS=$((($BLOCK_SIZE - 32) / (16 + 8 + 12 + 4)))
+	else
+		let MAX_ATTRS=$((($BLOCK_SIZE - 32) / (16 + 12 + 16 )))
+	fi
+	;;
 *)
 	# Assume max ~1 block of attrs
 	BLOCK_SIZE=`_get_block_size $TEST_DIR`
-- 
2.31.1

