[uml-user] can't trace kmem_cache_init with gdb

[Real N. <enj...@gm...>]

hi,

 I can't trace kmem_cache_init function with GDB, because gdb jump to wrong source line. Is there anyone know how to fix this issue?


[real@name tmp]$ tar -Jxf linux-3.12.6.tar.xz 
[real@name tmp]$ cd linux-3.12.6/
[real@name linux-3.12.6]$ make ARCH=um defconfig
  HOSTCC  scripts/basic/fixdep
  HOSTCC  scripts/kconfig/conf.o
  SHIPPED scripts/kconfig/zconf.tab.c
  SHIPPED scripts/kconfig/zconf.lex.c
  SHIPPED scripts/kconfig/zconf.hash.c
  HOSTCC  scripts/kconfig/zconf.tab.o
  HOSTLD  scripts/kconfig/conf
scripts/kconfig/conf --defconfig arch/x86/um/Kconfig
#
# configuration written to .config
#
[real@name linux-3.12.6]$ make ARCH=um linux 1>log.txt
arch/um/kernel/ptrace.c: In function ‘syscall_trace_enter’:
arch/um/kernel/ptrace.c:178:32: warning: ignoring return value of ‘tracehook_report_syscall_entry’, declared with attribute warn_unused_result [-Wunused-result]
  tracehook_report_syscall_entry(regs);
                                ^


[real@name linux-3.12.6]$ 
[real@name linux-3.12.6]$ cat gdb.txt 
file linux
handle SIGSEGV pass nostop noprint
handle SIGUSR1 pass nostop noprint

b kmem_cache_init
b create_boot_cache

r ubda=/home/real/uml/linux-3.12.6/Fedora20-x86-root_fs mem=256m
[real@name linux-3.12.6]$ 



[real@name linux-3.12.6]$ gdb -q  -x gdb.txt
Breakpoint 1 at 0x8050a57: file mm/slab.c, line 1512.
Breakpoint 2 at 0x804f7c5: file mm/slab_common.c, line 296.
Locating the bottom of the address space ... 0x10000
Locating the top of the address space ... 0xffffd000
Core dump limits :
	soft - 0
	hard - NONE
Checking that ptrace can change system call numbers...Detaching after fork from child process 16850.
OK
Checking syscall emulation patch for ptrace...Detaching after fork from child process 16851.
OK
Checking advanced syscall emulation patch for ptrace...Detaching after fork from child process 16852.
OK
Checking for tmpfs mount on /dev/shm.../dev/shm...nothing mounted on /dev/shm
Checking PROT_EXEC mmap in /tmp/...OK
Detaching after fork from child process 16853.
Checking for the skas3 patch in the host:
  - /proc/mm...not found: No such file or directory
  - PTRACE_FAULTINFO...Detaching after fork from child process 16854.
not found
  - PTRACE_LDT...Detaching after fork from child process 16855.
not found
UML running in SKAS0 mode

Breakpoint 1, kmem_cache_init () at mm/slab.c:1512
1512	{
Missing separate debuginfos, use: debuginfo-install glibc-2.17-20.fc19.i686
(gdb) c
Continuing.

Breakpoint 2, create_boot_cache (s=0x82bdaf8 <kmem_cache_boot>, name=0x82650f5 "kmem_cache", size=92, 
    flags=8192) at mm/slab_common.c:296
296	{
(gdb) n
299		s->name = name;
(gdb) 
296	{
(gdb) 
299		s->name = name;
(gdb) 
300		s->size = s->object_size = size;
(gdb) 
301		s->align = calculate_alignment(flags, ARCH_KMALLOC_MINALIGN, size);
(gdb) 
302		err = __kmem_cache_create(s, flags);
(gdb) s
__kmem_cache_create (cachep=0x82bdaf8 <kmem_cache_boot>, flags=8192) at mm/slab.c:2222
2222	{
(gdb) n
2226		size_t size = cachep->size;
(gdb) 
2251		if (size & (BYTES_PER_WORD - 1)) {
(gdb) 
2264		if (flags & SLAB_RED_ZONE) {
(gdb) 
2273		if (ralign < cachep->align) {
(gdb) 
2278			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
(gdb) 
2282		cachep->align = ralign;
(gdb) 
2284		if (slab_is_available())
(gdb) p slab_state
$1 = DOWN
(gdb) where
#0  __kmem_cache_create (cachep=0x82bdaf8 <kmem_cache_boot>, flags=8192) at mm/slab.c:2284
#1  0x0804f7f6 in create_boot_cache (s=0x82bdaf8 <kmem_cache_boot>, name=0x82650f5 "kmem_cache", 
    size=92, flags=8192) at mm/slab_common.c:302
#2  0x08050b1b in kmem_cache_init () at mm/slab.c:1559
#3  0x08049486 in mm_init () at init/main.c:468
#4  start_kernel () at init/main.c:532
#5  0x0804a6d8 in start_kernel_proc (unused=0x0) at arch/um/kernel/skas/process.c:46
#6  0x08059323 in new_thread_handler () at arch/um/kernel/process.c:140
#7  0x00000000 in ?? ()
(gdb) l mm/slab.c:2284
2279		/*
2280		 * 4) Store it.
2281		 */
2282		cachep->align = ralign;
2283	
2284		if (slab_is_available())
2285			gfp = GFP_KERNEL;
2286		else
2287			gfp = GFP_NOWAIT;
2288	

(gdb) p slab_state
$2 = DOWN
(gdb) n
2285			gfp = GFP_KERNEL;
(gdb) n
2289		setup_node_pointer(cachep);
(gdb) n
2285			gfp = GFP_KERNEL;


Start from 2289, gdb start jump to wrong source line. GDB should jump to 
"2287			gfp = GFP_NOWAIT;"
But it never hit it. And from here, gdb can't match the execuable UML kernel with source.


(gdb) n
2327		if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init &&



And I found that "file" command says the file format is different.
[real@name mm]$ pwd
/tmp/linux-3.12.6/mm
[real@name mm]$ file *.c | sort -k 2
percpu-vm.c:          ASCII text
filemap.c:            assembler source, ASCII text
memcontrol.c:         assembler source, ASCII text
ksm.c:                C source, ASCII text
mmap.c:               C source, ASCII text
rmap.c:               C source, ASCII text
slab.c:               C source, ASCII text
slob.c:               C source, ASCII text
slub.c:               C source, ASCII text

Why percpu-vm.c, filemap.c and memcontrol.c has different format?

thanks