Re: [RFA] gdbserver support for qCRC: (compare-sections)

[Pedro Alves <pedro@codesourcery.com>]

On Monday 22 March 2010 17:59:25, Michael Snyder wrote:
> 2010-03-19  Michael Snyder  <msnyder@vmware.com>
> 
>         * server.c (crc32): New function.
>         (handle_query): Add handling for 'qCRC:' request.
> 
> Index: server.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbserver/server.c,v
> retrieving revision 1.108
> diff -u -p -r1.108 server.c
> --- server.c    20 Jan 2010 22:55:38 -0000      1.108
> +++ server.c    22 Mar 2010 17:57:32 -0000
> @@ -788,6 +788,47 @@ handle_threads_qxfer (const char *annex,
>  
>  }
>  
> +/* Table used by the crc32 function to calcuate the checksum.  */
> +
> +static unsigned long crc32_table[256] =
> +{0, 0};
> +

I know you've copied this from remote.c, but, can't
we make this an `unsigned int' table?  longs are 64-bit
on most 64-bit archs, so this uses up double more memory
than needed.

> +/* Compute 32 bit CRC from inferior memory.
> +
> +   On success, return 32 bit CRC.
> +   On failure, return (unsigned long long) -1.  */
> +
> +static unsigned long long
> +crc32 (CORE_ADDR base, int len, unsigned int crc)

Can't say I'm thrilled by assuming
sizeof (long long) > sizeof (long), but I guess
it works on all targets gdbserver cares for presently.

> +{
> +  if (!crc32_table[1])
> +    {
> +      /* Initialize the CRC table and the decoding table.  */
> +      int i, j;
> +      unsigned int c;
> +
> +      for (i = 0; i < 256; i++)
> +       {
> +         for (c = i << 24, j = 8; j > 0; --j)
> +           c = c & 0x80000000 ? (c << 1) ^ 0x04c11db7 : (c << 1);
> +         crc32_table[i] = c;
> +       }
> +    }
> +
> +  while (len--)
> +    {
> +      unsigned char byte = 0;
> +
> +      /* Return failure if memory read fails.  */
> +      if (read_inferior_memory (base, &byte, 1) != 0)
> +       return (unsigned long long) -1;
> +
> +      crc = (crc << 8) ^ crc32_table[((crc >> 24) ^ byte) & 255];
> +      base++;
> +    }
> +  return (unsigned long long) crc;
> +}
> +
>  /* Handle all of the extended 'q' packets.  */
>  void
>  handle_query (char *own_buf, int packet_len, int *new_packet_len_p)
> @@ -1421,6 +1462,31 @@ handle_query (char *own_buf, int packet_
>        return;
>      }
>  
> +  if (strncmp ("qCRC:", own_buf, 5) == 0)
> +    {
> +      /* CRC check (compare-segment).  */
> +      char *comma;
> +      CORE_ADDR base = strtoul (own_buf + 5, &comma, 16);
> +      int len;
> +      unsigned long long crc;
> +

A `require_running' call is missing here.


> +      if (*comma++ != ',')
> +       {
> +         write_enn (own_buf);
> +         return;
> +       }
> +      len = strtoul (comma, NULL, 16);
> +      crc = crc32 (base, len, 0xffffffff);
> +      /* Check for memory failure.  */
> +      if (crc == (unsigned long long) -1)
> +       {
> +         write_enn (own_buf);
> +         return;
> +       }
> +      sprintf (own_buf, "C%lx", (unsigned long) crc);
> +      return;
> +    }
> +
>    /* Otherwise we didn't know what packet it was.  Say we didn't
>       understand it.  */
>    own_buf[0] = 0;

Otherwise, looks fine to me.

-- 
Pedro Alves