Re: Fatal BUG in store COSS in HEAD

The more correct fix is to move the cbdataReferenceDoneValid call up
above where the aqe is cleared.. one should not copy cbdata pointers
unless absolutely needed...

   int valid;

   ...

   valid = cbdataReferenceDoneValid(&aqe->callback_data, &cbdata);
 
   ...

   if (valid) {
        ...
   }

[full patch attached, applied to HEAD and 2.5]

Regards
Henrik

On Sunday 19 May 2002 13:54, Guido Serassio wrote:
> Hi Henrik,
>
> I'm working on port COSS on WIN32.
> I'have just found a fatal BUG in COSS:
> In the function a_file_callback() in async_io.c, the structure aqe
> is cleared before the call to cbdataReferenceValidDone(), so the
> cbdata pointer is invalid.
> The following can be a possible fix:
>
> Index: async_io.c
> ===================================================================
> RCS file: /cvsroot/squid/squid/src/fs/coss/async_io.c,v
> retrieving revision 1.3.26.1
> diff -u -p -r1.3.26.1 async_io.c
> --- async_io.c 17 Apr 2002 18:58:23 -0000 1.3.26.1
> +++ async_io.c 19 May 2002 11:53:07 -0000
> @@ -159,6 +159,7 @@ a_file_callback(async_queue_t * q)
> DWCB *wc;
> FREE *freefunc;
> void *cbdata;
> + void *callback_data;
> void *buf;
> int fd;
> async_queue_entry_t *aqe;
> @@ -186,6 +187,7 @@ a_file_callback(async_queue_t * q)
> buf = aqe->aq_e_buf;
> fd = aqe->aq_e_fd;
> type = aqe->aq_e_type;
> + callback_data = aqe->aq_e_callback_data;
>
> /* Free slot */
> bzero(aqe, sizeof(async_queue_entry_t));
> @@ -193,7 +195,7 @@ a_file_callback(async_queue_t * q)
> q->aq_numpending--;
>
> /* Callback */
> - if
> (cbdataReferenceValidDone(aqe->aq_e_callback_data, &cbdata)) {
> + if (cbdataReferenceValidDone(callback_data,
> &cbdata)) { if (type == AQ_ENTRY_READ)
> rc(fd, buf, retval, reterr, cbdata);
> if (type == AQ_ENTRY_WRITE)
>
> Regards
>
> Guido
>
>
>
> -
> =======================================================
> Serassio Guido
> Via Albenga, 11/4 10134 -
> Torino - ITALY E-mail: guido.serassio@serassio.it
> WWW: http://www.serassio.it