Fix issue with host header and canonicalURI

[jhiemstrawisc]

There were actually two issues that broke authenticated buckets. The first was that we weren't properly setting the host header. The second was that we weren't generating a valid canonical URI that gets signed and sent as a part of the Authorization header. The solution to both of these was to parse the hostUrl more carefully, and in only one spot.

Testing this is a bit of a challenging setup, so I'll try to loosely sketch this out here. This is from my notes, and it assumes you're running in a linux container as root. The container will need all the tools for building c++, as well as xrootd and the minio server binary. You'll also need a non-root user to run the xrootd process under. I'm also assuming you have some experience dealing with XRootD and all its funkiness.

Download & install minio server

Then, from terminal:
export MINIO_ROOT_USER="admin"
export MINIO_ROOT_PASSWORD="password"

The plugin operates under "virtual hosted" requests instead of amazon's old "path style" requests. To get minio working in this setup, you need to export the env var:
export MINIO_DOMAIN=`hostname`

And you also need to add this line to bottom of /etc/hosts:
172.17.0.XXXX      <bucket name>.<hostname>
where the '172.17.0.XXX' ip address should match the other IPs you see there, and where <bucket name> is the name of a
bucket you plan to create or have already created. I usually use 'test-bucket'

Next, launch the server to the path you want to export:
MINIO_ROOT_USER=$MINIO_ROOT_USER MINIO_ROOT_PASSWORD=$MINIO_ROOT_PASSWORD minio server `pwd`/test --console-address ":9001"

Now log into the server via browser to config test buckets: http://localhost:9001
user = admin
password = password

Then create a bucket (I usually call it test-bucket, explained above) and add a test file. Set the bucket to "private". Create
a user with "readwrite" permissions and configure an access/secret key for the user. Keep track of them, and write them
to `/etc/xrootd/access.key` and `/etc/xrootd/secret.key`

After building the plugin, you can test by using this xrootd config:

all.export  /
xrd.protocol http:8080 libXrdHttp.so

# Setting up S3 plugin
ofs.osslib <path to libXrdS3.so>
xrootd.async off

s3.service_name     s3.amazonaws.com
s3.region           us-east-1
# Next is whatever url your minio server gives you as its s3 endpoint, usually like below
s3.service_url      http://<hostname>:9000

# These should point to wherever you actually wrote the keys to
s3.access_key_file  /etc/xrootd/access.key
s3.secret_key_file  /etc/xrootd/secret.key

ofs.trace all
xrd.trace all -sched
http.trace all

Run the plugin under your non-root user: xrootd -c <configfile name>. You should get through all of the xrootd initialization.

Phew, you made it through all that! Give yourself a pat on the shoulder and do a quick stretch.

Finally, you should be able to get the test file from your s3 endpoint via your browser at:
http://localhost:1094/s3.amazonaws.com/us-east-1/<bucket name>/<object name>

Closes #7

[turetske]

Looks good except for wanting an added comment at the top of the function. Mostly to slowly update the documentation for this code as changes happen.

[github-actions]

[lint] _{reported by reviewdog 🐶}

Suggested change

	// Takes in the configured `s3.service_url` and uses the bucket/object requested
	// to generate the virtual host URL, as well as the canonical URI (which is the
	// path to the object).
	bool AmazonRequest::parseURL(const std::string &url, const std::string &bucket,
	const std::string &object, std::string &host,
	std::string &path) {
	auto i = url.find("://");
	if (i == std::string::npos) {
	return false;
	}
	// protocol = substring( url, 0, i );
	auto j = url.find("/", i + 3);
	if (j == std::string::npos) {
	host = bucket + "." + substring(url, i + 3);
	path = "/" + object;

[github-actions]

[lint] _{reported by reviewdog 🐶}

Suggested change

	return AWSv4Impl::canonicalizeQueryString( query_parameters );
	return AWSv4Impl::canonicalizeQueryString(query_parameters);