If a directory has a quota less than blockSize * numReplicas then you can't add a file to it, even if the file size is less than the quota. This is because FSDirectory#addBlock updates the count assuming at least one block is written in full. We don't know how much of the block will be written when addBlock is called and supporting such small quotas is not important so perhaps we should document this and log an error message instead of making small (blockSize * numReplicas) quotas work.
updateCount(inodes, inodes.length-1, 0, fileINode.getPreferredBlockSize()*fileINode.getReplication(), true);
You can reproduce with the following commands:
$ dd if=/dev/zero of=temp bs=1000 count=64
$ hadoop fs -mkdir /user/eli/dir
$ hdfs dfsadmin -setSpaceQuota 191M /user/eli/dir
$ hadoop fs -put temp /user/eli/dir # Causes DSQuotaExceededException