fts-solr: Returning 400 on searches; unescaped braces

Michael Welsh Duggan mwd at md5i.com
Mon Jan 16 19:52:24 UTC 2017


Michael Welsh Duggan <mwd at md5i.com> writes:

> Using Debian, dovecot-solr 1:2.2.26.0-4, and solr-tomcat 3.6.2+dfsg-9, I
> am getting 400 errors when doing searches.  Here is an example search
> query from dovecot that failed (captured with wireshark):
>
>   Frame 23: 338 bytes on wire (2704 bits), 338 bytes captured (2704 bits) on interface 0
>   Linux cooked capture
>   Internet Protocol Version 6, Src: ::1, Dst: ::1
>   Transmission Control Protocol, Src Port: 56860, Dst Port: 8080, Seq: 1, Ack: 1, Len: 250
>   Hypertext Transfer Protocol
>       GET /solr/select?fl=uid,score&rows=2664&sort=uid+asc&q={!lucene+q.op%3dAND}(hdr:test+OR+body:test)&fq=%2Bbox:6d5de009f991854df726000012cf7b9c+%2Buser:md5i HTTP/1.1\r\n
>       Host: localhost:8080\r\n
>       Date: Mon, 19 Dec 2016 00:25:56 GMT\r\n
>       Connection: Keep-Alive\r\n
>       \r\n
>       [Full request URI: http://localhost:8080/solr/select?fl=uid,score&rows=2664&sort=uid+asc&q={!lucene+q.op%3dAND}(hdr:test+OR+body:test)&fq=%2Bbox:6d5de009f991854df726000012cf7b9c+%2Buser:md5i]
>       [HTTP request 1/1]
>       [Response in frame: 25]
>
> Here is the same query from firefox, which succeeds:
>
>   Frame 66: 646 bytes on wire (5168 bits), 646 bytes captured (5168 bits) on interface 0
>   Linux cooked capture
>   Internet Protocol Version 6, Src: ::1, Dst: ::1
>   Transmission Control Protocol, Src Port: 56862, Dst Port: 8080, Seq: 1, Ack: 1, Len: 558
>   Hypertext Transfer Protocol
>       GET /solr/select?fl=uid,score&rows=2664&sort=uid+asc&q=%7B!lucene+q.op%3DAND%7D(hdr:test+OR+body:test)&fq=%2Bbox:6d5de009f991854df726000012cf7b9c+%2Buser:md5i HTTP/1.1\r\n
>       Host: localhost:8080\r\n
>       Connection: keep-alive\r\n
>       Cache-Control: max-age=0\r\n
>       Upgrade-Insecure-Requests: 1\r\n
>       User-Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.75 Safari/537.36\r\n
>       Accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8\r\n
>       DNT: 1\r\n
>       Accept-Encoding: gzip, deflate, sdch, br\r\n
>       Accept-Language: en-US,en;q=0.8\r\n
>       \r\n
>       [Full request URI: http://localhost:8080/solr/select?fl=uid,score&rows=2664&sort=uid+asc&q=%7B!lucene+q.op%3DAND%7D(hdr:test+OR+body:test)&fq=%2Bbox:6d5de009f991854df726000012cf7b9c+%2Buser:md5i]
>       [HTTP request 1/1]
>       [Response in frame: 86]
>
>
> The salient difference seems to be the encoding of the braces.  Indeed
> in the tomcat 8 logs, I find the following which seems to corroborate
> my hypothesis:
>
>  java.lang.IllegalArgumentException: Invalid character found in the request target. The valid characters are defined in RFC 7230 and RFC 3986
> 	at org.apache.coyote.http11.Http11InputBuffer.parseRequestLine(Http11InputBuffer.java:467)
> 	at org.apache.coyote.http11.Http11Processor.service(Http11Processor.java:667)
> 	at org.apache.coyote.AbstractProcessorLight.process(AbstractProcessorLight.java:66)
> 	at org.apache.coyote.AbstractProtocol$ConnectionHandler.process(AbstractProtocol.java:789)
> 	at org.apache.tomcat.util.net.NioEndpoint$SocketProcessor.doRun(NioEndpoint.java:1437)
> 	at org.apache.tomcat.util.net.SocketProcessorBase.run(SocketProcessorBase.java:49)
> 	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
> 	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
> 	at org.apache.tomcat.util.threads.TaskThread$WrappingRunnable.run(TaskThread.java:61)
> 	at java.lang.Thread.run(Thread.java:745)
>
> Indeed the braces are listed in fts-backend-solr.c as part of
> solr_escape_chars, so don't know how the braces are making it through
> unencoded.

I have attached a patch which solves this problem.  I initially tried
changing http_url_escape_param() to include braces, but this did not
solve the problem.  I have to guess that the {!lucene+q.op=AND} bit does
not travel through this function.  So I just changed the braces in the
lines where they were introduced into their encoded values.  Since the
equals-sign was already encoded this way there, it seemed to make sense.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: escape-braces.patch
Type: text/x-diff
Size: 831 bytes
Desc: not available
URL: <http://dovecot.org/pipermail/dovecot/attachments/20170116/ec00edb6/attachment.bin>
-------------- next part --------------

-- 
Michael Welsh Duggan
(md5i at md5i.com)


More information about the dovecot mailing list