[Samba] Printing eats my CPU

[Franz Sirl]

At 09:54 29.07.2002, Franz Sirl wrote:
>Note: Resent to list only without attachment
>
>At 16:41 08.07.2002, Gerald Carter wrote:
>>On Tue, 2 Jul 2002, Martin Zielinski wrote:
>>
>> > Hi all,
>> > I still have several problems with the printing system.
>> >
>> > The most critical is, that some drivers seem to cause an infinite RPC 
>> loop.
>> > I have a KYOCERA Mita KM 4230 KX driver installed. In this moment I 
>> can see
>> > SPOOLSS_GET_PRINTER and SPOOLSS_GET_PRINTERDATA requests and
>> > RPC responses on the net. Stracing the smbd process looks like this:
>>
>>How many printers do you have on the server?  Just looks like normal
>>tdb lookups.
>
>I got the same problem here with the Kyocera FS-9000 KX driver rev 1.7 and 
>a single printer share on the server. Occasionally printing wouldn't work 
>anymore and some smbd processes did cause high CPU load, so far the only 
>measure I found was to delete and recreate all the printing TDB's.
>
>Today it happened again for one colleague and he answered "I fiddled with 
>the printer settings, playing with the prolog/epilog sequences". This 
>caught my attention and I tried the same from my Win2ksp1 PC. I just 
>enabled and disabled the prolog/epilog setting in the driver once, pressed 
>OK, and voila! My smbd process did go to high CPU load!
>
>In the logs lines like
>
>[2002/07/26 15:57:57, 0] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7935)
>   talloc_realloc failed to allocate more memory for data!
>
>arrived quickly. I looked in the source and then added this patch:
>
>-                       DEBUG(0,("talloc_realloc failed to allocate more 
>memory for data!\n"));
>+                       DEBUG(0,("talloc_realloc failed to allocate more 
>memory [data_len=%d] [add_len=%d] for data!\n", data_len, add_l
>
>and then I did a level 10 trace:
>
>[2002/07/26 15:57:57, 9] printing/nt_printing.c:get_a_printer_2(2701)
>   Unpacked printer [fs9000] name [\\linuxpc1\fs9000] running driver 
> [Kyocera FS-9000 KX]
>[2002/07/26 15:57:57, 10] printing/nt_printing.c:get_a_printer(3302)
>   get_a_printer: [fs9000] level 2 returning WERR_OK
>[2002/07/26 15:57:57, 10] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7919)
>   retrieved value number [0] [Setup]
>[2002/07/26 15:57:57, 10] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7919)
>   retrieved value number [1] [Render1]
>[2002/07/26 15:57:57, 10] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7919)
>   retrieved value number [2] [PDL1]
>[2002/07/26 15:57:57, 10] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7919)
>   retrieved value number [3] [PDL4]
>[2002/07/26 15:57:57, 10] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7919)
>   retrieved value number [4] [BinReassign]
>[2002/07/26 15:57:57, 10] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7919)
>   retrieved value number [5] [IsUserSetChange]
>[2002/07/26 15:57:57, 10] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7919)
>   retrieved value number [6] [TempKcPrologEpilogData]
>[2002/07/26 15:57:57, 0] 
>rpc_server/srv_spoolss_nt.c:_spoolss_enumprinterdataex(7935)
>   talloc_realloc failed to allocate more memory [data_len=0] [add_len=0] 
> for data!
>
>Aha, trying to allocate 0 bytes failed! I have no idea how this RPC stuff 
>works, but my guess is that my enabling/of the prolog/epilog sequences 
>(TempKcPrologEpilogData?) in the printer driver settings created the 
>entry, but since I disabled it immediately again, it wasn't deleted, but 
>set to 0 length.

Looking at other places in the samba source, data_len==0 seems to be a 
perfectly valid condition, so I tried the appended patch and all is fine now!

Franz.

         * rpc_server/srv_spoolss_nt.c (_spoolss_enumprinterdataex): Don't 
try to allocate 0 bytes.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: samba-highload.patch
Type: application/octet-stream
Size: 1301 bytes
Desc: not available
Url : http://lists.samba.org/archive/samba/attachments/20020729/5fcdbef9/samba-highload.obj