Re: SSL information view - Mailing list pgsql-hackers

From Magnus Hagander
Subject Re: SSL information view
Date
Msg-id CABUevEyLMvoTn=oWmh0POYKkip=La0+gN7=JEO0C7Sd_rT6JzQ@mail.gmail.com
Whole thread Raw
In response to Re: SSL information view  (Heikki Linnakangas <hlinnakangas@vmware.com>)
Responses Re: SSL information view
List pgsql-hackers
On Wed, Dec 17, 2014 at 9:19 PM, Heikki Linnakangas <hlinnakangas@vmware.com> wrote:
On 11/19/2014 02:36 PM, Magnus Hagander wrote:
+       /* Create or attach to the shared SSL status buffers */
+       size = mul_size(NAMEDATALEN, MaxBackends);
+       BackendSslVersionBuffer = (char *)
+               ShmemInitStruct("Backend SSL Version Buffer", size, &found);
+
+       if (!found)
+       {
+               MemSet(BackendSslVersionBuffer, 0, size);
+
+               /* Initialize st_ssl_version pointers. */
+               buffer = BackendSslVersionBuffer;
+               for (i = 0; i < MaxBackends; i++)
+               {
+                       BackendStatusArray[i].st_ssl_version = buffer;
+                       buffer += NAMEDATALEN;
+               }
+       }
+
+       size = mul_size(NAMEDATALEN, MaxBackends);
+       BackendSslCipherBuffer = (char *)
+               ShmemInitStruct("Backend SSL Cipher Buffer", size, &found);
+
+       if (!found)
+       {
+               MemSet(BackendSslCipherBuffer, 0, size);
+
+               /* Initialize st_ssl_cipher pointers. */
+               buffer = BackendSslCipherBuffer;
+               for (i = 0; i < MaxBackends; i++)
+               {
+                       BackendStatusArray[i].st_ssl_cipher = buffer;
+                       buffer += NAMEDATALEN;
+               }
+       }
+
+       size = mul_size(NAMEDATALEN, MaxBackends);
+       BackendSslClientDNBuffer = (char *)
+               ShmemInitStruct("Backend SSL Client DN Buffer", size, &found);
+
+       if (!found)
+       {
+               MemSet(BackendSslClientDNBuffer, 0, size);
+
+               /* Initialize st_ssl_clientdn pointers. */
+               buffer = BackendSslClientDNBuffer;
+               for (i = 0; i < MaxBackends; i++)
+               {
+                       BackendStatusArray[i].st_ssl_clientdn = buffer;
+                       buffer += NAMEDATALEN;
+               }
+       }

This pattern gets a bit tedious. We do that already for application_names, client hostnames, and activity status but this adds three more such strings. Why are these not just regular char arrays in PgBackendStatus struct, anyway? The activity status is not, because its size is configurable with the pgstat_track_activity_query_size GUC, but all those other things are fixed-size.

Also, it would be nice if you didn't allocate the memory for all those SSL strings, when SSL is disabled altogether. Perhaps put the SSL-related information into a separate struct:

struct
{
        /* Information about SSL connection */
        int             st_ssl_bits;
        bool            st_ssl_compression;
        char            st_ssl_version[NAMEDATALEN];  /* MUST be null-terminated */
        char            st_ssl_cipher[NAMEDATALEN];   /* MUST be null-terminated */
        char            st_ssl_clientdn[NAMEDATALEN]; /* MUST be null-terminated */
} PgBackendSSLStatus;

Those structs could be allocated like you allocate the string buffers now, with a pointer to that struct from PgBackendStatus. When SSL is disabled, the structs are not allocated and the pointers in PgBackendStatus structs are NULL.


Finally, I found time to do this. PFA a new version of this patch.

It takes into account the changes suggested by Heikki and Alex (minus the renaming of fields - I think that's a separate thing to do, and we should stick to existing naming conventions for now - but I changed the order of the fields). Also the documentation changes suggested by Peter (but still not the contrib/sslinfo part, as that should be a separate patch - but I can look at that once we agree on this one). And resolves the inevitable oid conflict for a patch that's been delayed that long. 

--
Attachment

pgsql-hackers by date:

Previous
From: David Rowley
Date:
Subject: Re: Parallel Seq Scan
Next
From: Andres Freund
Date:
Subject: Re: NOT NULL markings for BKI columns