Jeff,
If you have a patch against 2.18.1 for the pipe problem, I will apply it.
Again, perhaps a strtst test case could be added as well.
--brian
On Fri, 29 Oct 2004, Jeff Goldszer wrote:
>
> Gentlemen,
>
> Below are resolutions to some of the issues I reported in earlier
> emails. I have emailed a patch to Dave Grothe which addresses items 2
> and 3. Item 1 is a false alarm.
>
> 1) lis_freezestr and lis_freezestr do indeed accept any queues
> other the stream head pair. I misread the code. All queue's have
> pointers to the stream head (q_str).
>
> 2) The changed I made two years ago is still maintained in
> LiS-2.18 . The problem is a there are places in the code that freemsg
> on an M_PASSFP message. The file descriptor in a M_PASSFP message
> should be closed before as well as freeing the mblk. The following
> test program can fail if file descriptor is located on a[0] streams
> read put function queue (sd_rput_hd) when a[0] is closed. The failure
> is that the file descriptor b[0] will not be freed and LiS can not be
> unloaded.
>
> #include <unistd.h>
>
> #include <stropts.h>
>
> main()
>
> {
>
> int a[2];
>
> int b[2];
>
> pipe(a);
>
> pipe(b);
>
> ioctl(a[0], I_SENDFD, b[0]);
>
> close(b[0]);
>
> close(a[1]);
>
> }
>
> 3) There is a synchronization problem between pipes. One such
> issue resulted in an Oops. The problem is pipe disassemble queues in
> both directions. There is an unprotected critical area. I.e., one
> process is calling putmsg on one side of a pipe, while another process
> is closing the other side of the pipe. One process is closing one side
> of the pipe while another process is closing the other side of the
> pipe.
>
>
> Below is the an oops cause by issue 3.
>
> Agent: TERMINATED
>
> /u<4s>r//usrsrc//sLircS-/L2.i1S-7/2.h1e7a/d/headed/h.cea #d.4c17
> #0:4 1Q7oue:
>
> uqe uemauegi mc engiumc benrum biser 0xis93 03x9903d3d90,0 sddh,ou
> sldho bule
>
> d0 bx9e 330x990033dc90
>
> dc 0
>
> Unable to handle kernel NULL pointer dereference at virtual address
> 00000090
>
> printing eip:
>
> f8a0d9e9
>
> *pde = 00000000
>
> Oops: 0002
>
> pl2303 usbserial localdrv snamux lu62mux sessmux sdlcmux
> streams-adpa streams-cxio
>
> streams-copier placxp slmdrv streams bcm5700 st loop keybdev
> mousedev hid i
>
> CPU: 2
>
> EIP: 00b0:[<f8a0d9e9>] Tainted: PF
>
> EFLAGS: 00010046
>
> EIP is at lis_rw_write_lock_irqsave_fcn [streams] 0x11
> (2.4.21-15.0.4.EL-1/i686)
>
> eax: 00000090 ebx: 00000000 ecx: 00000058 edx:
> 00000090
>
> esi: 00000058 edi: eb81b960 ebp: e4b29e40 esp:
> eb429e18
>
> do: 0068 es: 0068 ss: 0068
>
> Process stopcdm ( pid: 21360, stackpage = e4b29000 )
>
> Stack: f89f7158 00000058 e4b29e40 f8a0ef4d 0000104b
> eb81b960 f8a0ef4d 0000104a
>
> eb81b960 eb81b960 00000246 00000000 00000000
> eaad2100 eae45d44 f89f73bc
>
> eaad2100 c312c100 eae45d44 00000000 00000000
> 00000001 eae45d44 f89f427b
>
>
> Call Trace: [<f89f7158>] check_for_wantenable [streams] 0x94 (
> 0xe4b29e18 )
>
> [<f8a0ef4d>] .rodata.str1.1 [streams] 0x2fd (0xe4b29e24 )
>
> [<f8a0ef4d>] .rodata.str1.1 [streams] 0x2fd ( 0xe4b29e30 )
>
> [<f89f73bc>] lis_unfreezestr [streams] 0x64 ( 0xe4b29e54 )
>
> [<f89f427b>] lis_qdetach [streams] 0x54b ( 0xe4b29e74 )
>
> [<f8a0ef4d>] .rodata.str1.1 [streams] 0x2fd ( 0xe4b29ea4 )
>
> [<f89f2e21>] lis_dismantle [streams] 0x61 ( 0xe4b29ed0 )
>
> [<f89fcfe3>] close_action [streams] 0x197 ( 0xe4b29eec )
>
> [<f89fd5a0>] lis_doclose [streams] 0x328 ( 0xe4b29f18 )
>
> [<f89fd7a0>] lis_strclose [streams] 0x178 ( 0xe4b29f44 )
>
> [<c015507b>] __fput [kernel] 0xe4 ( 0xe4b29f78 )
>
> [<c0153440>] filp_close [kernel] 0x92 ( 0xe4b29f94 )
>
> [<c01534e2>] sys_close [kernel] 0x66 ( 0xe4b29fb0 )
>
> Code: f0 81 28 00 00 00 01 0f 85 a8 06 00
> 00 b8 00 ff 21 e0
>
> Kernel panic: Fatal exception
>
> Jeff Goldszer
>
> Senior Software Engineer
>
> Alebra Technologies
>
> 9915 66th Avenue
>
> Apartment 2C
>
> Rego Park, NY 11374
>
> Phone: 718-606-8535
>
> FAX: 718-606-8536
--
Brian F. G. Bidulock � The reasonable man adapts himself to the �
[EMAIL PROTECTED] � world; the unreasonable one persists in �
http://www.openss7.org/ � trying to adapt the world to himself. �
� Therefore all progress depends on the �
� unreasonable man. -- George Bernard Shaw �
_______________________________________________
Linux-streams mailing list
[email protected]
http://gsyc.escet.urjc.es/mailman/listinfo/linux-streams
