tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jason Smith <jsm...@infotrustgroup.com>
Subject Help with a Tomcat issue???
Date Mon, 06 Apr 2009 18:27:06 GMT

Sorry for the spam if you are on both newsgroups.  I have been asked to move this to the
user's group rather than the developers group.  I still think it's a developer issue, but
I am more interested in actually finding answers than arguing the point.

I am using a standard UrlHttpConnection to connect to Tomcat.  When I use setChunkedStreamingMode,
no matter what the setting, I start getting bad method names in my servlet.

HttpURLConnection conn = (HttpURLConnection)url.openConnection();
conn.setChunkedStreamingMode(1024);

I am seeing this in my override of the .service() method in the Servlet, which is the earliest
point available to me if I stay within the confines of the spec.  The method name comes in
as '0\n\nPOST'.  I've tracked this through a bunch of Tomcat classes, working on the 5.5.27
code base.  

I have intercepted the socket InputStream (in Http11BaseProtocol) and looked at the data coming
in.  It is what I expect ('POST .').  Yes, let me repeat that, the data coming in from the
socket is good.  I've visually inspected it, and it matches the spec.

When I get into the InternalInputBuffer, I am seeing there is data left over at the end of
one stream ('0\n\n') which gets appended to the beginning of the next request because of the
code in .nextRequest().  It almost looks like someone downstream is not consuming the very
last byte in the buffer.

But then why the heck does .nextRequest copy it into the new buffer for the next request? 
This makes no sense to me.

So I have verified that the data is coming in right.  And at the very first entry point into
the servlet, the method name is messed up.  

This appears to be happening in Tomcat, and has been replicated on a number of machines.

So PLEASE, can ANYONE tell me what .nextRequest() is doing?  It says that *** All bytes of
the current request should have been already consumed.***  That is not what I am seeing.
And this part of the code isn't dealing with corrupt data very elegantly (it assumes that
everything is perfect).

Is this a problem with downstream code, which has a responsibility to consume every byte in
the input stream? Or is it bad code in nextRequest(), which really should not be taking the
tail of one request and appending it to the next?  Or is it something else?

Thanks!  See below for more info.



From: Jason Smith 
Sent: Monday, April 06, 2009 11:08 AM
To: 'dev@tomcat.apache.org'
Subject: RE: Help with a Tomcat bug.

More info.  In InternalInputBuffer.nextRequest(), I noticed there is code to pull remaining
bytes into the current buffer before switching.  

    /**
     * End processing of current HTTP request.
     * Note: All bytes of the current request should have been already 
     * consumed. This method only resets all the pointers so that we are ready
     * to parse the next HTTP request.
     */
    public void nextRequest()
        throws IOException {

        // Recycle Request object
        request.recycle();

        // Determine the header buffer used for next request
        byte[] newHeaderBuf = null;
        if (buf == headerBuffer1) {
            newHeaderBuf = headerBuffer2;
        } else {
            newHeaderBuf = headerBuffer1;
        }

        // Copy leftover bytes from buf to newHeaderBuf
        System.arraycopy(buf, pos, newHeaderBuf, 0, lastValid - pos);
        if(lastValid-pos > 0)
        {
            System.out.println("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@");
            System.out.println("'" + new String(Arrays.copyOf(newHeaderBuf, lastValid
- pos), "US-ASCII") + "'");
        }

        // Swap buffers
        buf = newHeaderBuf;

        // Recycle filters
        for (int i = 0; i <= lastActiveFilter; i++) {
            activeFilters[i].recycle();
        }

        // Reset pointers
        lastValid = lastValid - pos;
        pos = 0;
        lastActiveFilter = -1;
        parsingHeader = true;
        swallowInput = true;

    }

I am seeing something like this at one point:

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
'POST /dh/services/jmap/__exists__ HTTP/1.1

But I am also seeing this where this problem is cropping up:

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
'0

'

Anyone got any ideas on how to fix this?  Data from one POST is being carried over to the
next POST!!!!!

From: Jason Smith 
Sent: Monday, April 06, 2009 10:16 AM
To: 'dev@tomcat.apache.org'
Subject: Help with a Tomcat bug.

When using .setChunkedStreamingMode(.) from the client, I was getting back an invalid method
name in my servlet.  Specifically, in the overridden service() method, the request.getMethod()
was returning '0\n\nPOST'.  

I've tracked this all the way into  org.apache.coyote.http11.InternalInputBuffer.  

In .parseRequestLine, the first thing it does is consume leading CRs and LFs.  Well, the
first line I am getting is '0\n'.  So it won't consume that line.

The next step parses to the next SPACE character.  So it picks up the 0, the CRs and LFs,
all the way to the end of POST.  

The bottom line is that at this point, in this method, the HTTP method name is already messed
up.  

Should this be fixed in this method, or is there a better place?

One quick fix:

byte chr = 0;
        do {

            // Read new bytes if needed
            if (pos >= lastValid) {
                if (!fill())
                    throw new EOFException(sm.getString("iib.eof.error"));
            }

            chr = buf[pos++];

        } while ((chr == Constants.CR) || (chr == Constants.LF) || (chr == '0'));


I simply check for the '0' character as well.  This is a bit of a hack, but I don't know
the code well enough to know if the leading '0' (which I believe is the last line from a previous
chunked POST) is supposed to be there or not.

Any help would be appreciated.  

Tomcat 5.5.27, Java 6u13.

Jason Smith
Software Engineer 
InfoTrust Group, Inc.




---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org


Mime
View raw message