Problems with l4linux

Valery V. Sedletski _valerius at
Thu Oct 23 10:54:08 CEST 2008

On Wed, 22 Oct 2008 19:12:59 +0200, Adam Lackorzynski wrote:

>Ok. There's an option in the Fiasco configuration to let you set IOPL3
>(and thus lets you do cli/sti). This is an extra option because if you
>enable it Linux can stop the machine. Anyway, that sould fix the X

Now I switched on Fiasco option about IOPL 3 and an error with cli disappeared,
but another problem -- GPF somewhere in Fiasco. The serial log is like this:

loader  | vmlinuz26: Starting application using
loader  | vmlinuz26,#f: Loading binary
rtc     | Date:23.10.2008 Time:08:17:15
loader  | vmlinuz26,#f: Loading ldso
loader  | vmlinuz26,#f: Starting at 000138e0 via 0000cbe8
l4lx    | ======> L4Linux 2.6 starting... <========
l4lx    | Linux version 2.6.26-l4-svn120-dirty2 (root at localhost) (gcc version 4
l4lx    : .2.2 20071128 (prerelease) (4.2.2-3mdv2008.0)) #7 Thu Oct 23 21:07:12
l4lx    :  PETST 2008
l4lx    | Binary name: vmlinuz26
l4lx    | Linux kernel command line (8 args): mem=100M video=l4fb:refreshsleep:
l4lx    : 200 l4fb.nograb=1 load_ramdisk=1 ramdisk_size=40960 root=/dev/ram l4e
l4lx    : nv_rd=(nd)/tftpboot/drops/ramdisk/drops-fp.rd panicblink=0
l4lx    | Image: 00400000 - 007b9000 [3812 KiB].
l4lx    | Areas: Text:     00400000 - 006c3000 [2828kB] (a bit longer)
l4lx    |        Data:     006c3000 - 006e46d8 [133kB]
l4lx    |        Initdata: 006e8000 - 00726000 [248kB]
l4lx    |        BSS:      00728000 - 007b8bb0 [578kB]
l4lx    | l4lx_thread_create: Created thread 0f.03 (tamer0)
l4lx    | Tamer0 is 0f.03
l4lx    | Using tamed mode.
ROOT: Sending all ports (for cli/sti) to task #0d
l4lx    | Got 65536 out of 65536 I/O ports
l4lx    | Running at IOPL 3
l4lx    | Connecting to l4io server.
l4lx    | l4lx_thread_create: Created thread 0f.04 (cpu0)
l4lx    | main thread will be 0f.04
l4lx    | l4env_register_pointer_section: addr = 006e6000 size = 266240
l4lx    |      sec-w-init: virt: 0x006e6000 to 0x00726fff [260 KiB]
l4lx    |      sec-w-init: Number of physical regions: 1, 266240 Bytes
l4lx    |      sec-w-init: 1: Phys: 0x012a0000 to 0x012e1000, Size:   266240
l4lx    | l4env_linux_startup thread 4.
l4lx    | main thread: received startup message.
l4lx    | Main thread running, waiting...
l4lx    | setup_l4env_memory: Forcing superpages for main memory
l4lx    | Main memory size: 100MB
l4lx    |     Main memory: virt: 0x00800000 to 0x06bfffff [102400 KiB]
l4lx    |     Main memory: Number of physical regions: 1, 104857600 Bytes
l4lx    |     Main memory: 1: Phys: 0x06400000 to 0x0c800000, Size: 104857600
l4lx    | Filling lower ptabs...
l4lx    | mainmem = 800000
l4lx    | Done (1060 entries).
l4lx    | l4env_register_pointer_section: addr = 00728000 size = 593920
l4lx    |             end: virt: 0x00728000 to 0x007b8fff [580 KiB]
l4lx    |             end: Number of physical regions: 1, 593920 Bytes
l4lx    |             end: 1: Phys: 0x006f2000 to 0x00783000, Size:   593920
l4lx    | l4env_rd_path: (nd)/tftpboot/drops/ramdisk/drops-fp.rd
l4lx    | Loading: (nd)/tftpboot/drops/ramdisk/drops-fp.rd
l4lx    | INITRD: Size of RAMdisk is 40960KiB
l4lx    | RAMdisk from 10000000 to 12800000 [40960KiB]
l4lx    | l4lx_thread_create: Created thread 0f.05 (timer.i0)

    ---------------------------------------------------------IP: f003622b      
    --General Protection
[l4lx.main] (f.02) jdb: 

-- TRAP 000d at 0xf003622b. Here is a disassembly of Fiasco kernel:

f0036224 <in_syscall>:
f0036224:       59                      pop    %ecx
f0036225:       5a                      pop    %edx
f0036226:       5e                      pop    %esi
f0036227:       5f                      pop    %edi
f0036228:       5b                      pop    %ebx
f0036229:       5d                      pop    %ebp
f003622a:       58                      pop    %eax
f003622b:       cf                      iret
f003622c:       8d 74 26 00             lea    0x0(%esi),%esi

f0036230 <entry_sys_id_nearest>:
f0036230:       50                      push   %eax
f0036231:       b8 c4 37 05 f0          mov    $0xf00537c4,%eax
f0036236:       eb c8                   jmp    f0036200 <all_syscalls>

f0036238 <entry_sys_fpage_unmap>:
f0036238:       50                      push   %eax
f0036239:       b8 c8 37 05 f0          mov    $0xf00537c8,%eax
f003623e:       eb c0                   jmp    f0036200 <all_syscalls>

So, this is iret instruction somewhere on return from a syscall, if I understood correctly.
Maybe, my Fiasco config is wrong?


 * Automatically generated C config: don't edit
 * Fiasco kernel version: SVN
 * Thu Oct 23 21:11:17 2008
#define CONFIG_IA32_TARGET "Intel Pentium"
#define CONFIG_HOST_CC "gcc"
#define CONFIG_IO_PROT 1
#define CONFIG_BIT32 1
#define CONFIG_LABEL ""
#define CONFIG_XARCH "ia32"
#define CONFIG_CC "gcc"
#define CONFIG_PF_PC 1
#define CONFIG_HOST_CXX "g++"
#define CONFIG_ABI_V2 1
#define CONFIG_ABI "v2"
#define CONFIG_JDB 1
#define CONFIG_IA32 1
#define CONFIG_IA32_586 1
#define CONFIG_CXX "g++"

>(My X version also tries to get ports 0-1024 (or similar)
>before doing the iopl thing but AFAIK this has only been added shortly.)

-- so, in normal situation, it should not be (requesting access to some ports before doing iopl(3)), 
but if it takes the place, then for it to work, the Fiasco option mentioned must be enabled. But in
normal situation, this option is not needed?

And the question: what does cli instruction do? I guess this must trigger an I/O page fault for I/O
pages to be mapped? Am I right? 

>> >> PS: When I did "modprobe serial_cs" udev created ttyS0..ttyS3 devices but still no comm port access. So, access to video
>> >> and comm-ports don't work but network and disks are working. Why could it be?
>> >
>> >Sorry, no idea with the card. It doesn't say anything strange in dmesg?
>> >
>> Only that serial_cs module prints the message that it could not access
>> high memory. I hope this error disappear after linux will get I/O
>> privileges
>I guess not. Which address? Anyway, I guess it should ioremap those...

According to scanpci, Cardbus controller and serial (modem) card use these addresses:

pci bus 0x0006 cardnum 0x09 function 0x00: vendor 0x104c device 0x8031
 Texas Instruments PCIxx21/x515 Cardbus Controller
  STATUS    0x0210  COMMAND 0x0007
  CLASS     0x06 0x07 0x00  REVISION 0x00
  BIST      0x00  HEADER 0x82  LATENCY 0xa8  CACHE 0x08
  BASE0     0xb8008000  addr 0xb8008000  MEM
  BASE1     0x020000a0  addr 0x020000a0  MEM
  BASE2     0xb00a0706  addr 0xb00a0700  MEM
  BASE3     0x88000000  addr 0x88000000  MEM
  BASE4     0x8bfff000  addr 0x8bfff000  MEM
  BASE5     0x90000000  addr 0x90000000  MEM
  BASEROM   0x000044fc  addr 0x00000000  not-decode-enabled
  MAX_LAT   0x05  MIN_GNT 0xc0  INT_PIN 0x01  INT_LINE 0x0a
  BYTE_0    0xc0  BYTE_1  0x17  BYTE_2  0x07  BYTE_3  0x30

according to /proc/iomem:

  88000000-8bffffff : PCI CardBus #07
90000000-93ffffff : PCI CardBus #07
    b8008000-b8008fff : yenta_socket
  b8010000-b8010fff : pcmcia_socket0

And also it uses these I/O ports (according /proc/ioport):

03f8-0407 : pcmcia_socket0
  03f8-03ff : serial
  0400-0407 : serial
  4400-44ff : PCI CardBus #07
  4800-48ff : PCI CardBus #07

0x3f8-0x3ff is /dev/ttyS0
0x400-0x407 is /dev/ttyS1


More information about the l4-hackers mailing list