Unable to start 7.2.5 crash with SLES15 coredump

[alex1284B]

I've got
crash-7.2.5: page excluded: kernel virtual address: ffff880825d47000 type: "memory section"

7.2.1 crash works fine with the same coredump/vmlinux. Both versions were builded with the same gcc/gdb.
build_target: X86_64
build_version: 7.2.5
compiler version: gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-17)
SLES kernel is 4.12.14-25.22 x86_64 SMP
I've checked 7.2.2 and it doesn't work also with the same error msg.

[crash-utility]

----- Original Message -----

I've got crash-7.2.5: page excluded: kernel virtual address: ffff880825d47000 type: "memory section" 7.2.1 crash works fine with the same coredump/vmlinux. Both versions were builded with the same gcc/gdb. build_target: X86_64 build_version: 7.2.5 compiler version: gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-17) SLES kernel is 4.12.14-25.22 x86_64 SMP I've checked 7.2.2 and it doesn't work also with the same error msg. -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: #28

Looks like it might associated with these to 7.2.2 patches: - As the first step in optimizing the is_page_ptr() function, save the maximum SPARSEMEM section number during initialization, and use it as the topmost delimeter in subsequent mem_section searches. Also allow for per-architecture machdep->is_page_ptr() plugin functions. ([email protected]) - Implemented the x86_64 machdep->is_page_ptr() plugin function. If the kernel is configured with CONFIG_SPARSEMEM_VMEMMAP, the plugin function optimizes the mem_section search, reducing the computation effort and time consumed by commands that repeatedly call the is_page_ptr() function on large-memory systems. ([email protected]) You could try commenting out line 161 in x86_64.c: 157 switch (when) 158 { 159 case SETUP_ENV: 160 machdep->process_elf_notes = x86_process_elf_notes; 161 machdep->is_page_ptr = x86_64_is_page_ptr; 162 break; 163 case PRE_SYMTAB: Dave

[alex1284B]

I've tried, the same error.

[pandafromspace]

I can confirm the same issue. The regression appeared with commit d586679. Removing the line "machdep->is_page_ptr = x86_64_is_page_ptr;" after having checked out commit 4141373 does not help, crash fails with the same "memory section" error message.

The last "memory section" references with the modified build are as follows:

Breakpoint 1, read_mem_section (addr=18446683600716103656) at memory.c:17005
17005 if ((addr == 0) || !IS_KVADDR(addr))
#0 read_mem_section (addr=18446683600716103656) at memory.c:17005
#1 0x00005555556ad30c in valid_section (addr=18446683600716103656) at memory.c:17066
#2 0x00005555556ad4a6 in valid_section_nr (nr=487456) at memory.c:17112
#3 0x00005555556ad654 in dump_mem_sections (initialize=1) at memory.c:17155
#4 0x00005555556aa88c in dump_memory_nodes (initialize=1) at memory.c:16309
#5 0x00005555556ab947 in node_table_init () at memory.c:16536
#6 0x000055555567322c in vm_init () at memory.c:1093
#7 0x0000555555656a98 in main_loop () at main.c:770
#8 0x0000555555909bb3 in captured_command_loop (data=data@entry=0x0) at main.c:258
#9 0x00005555559086fa in catch_errors (func=func@entry=0x555555909ba0 <captured_command_loop>, func_args=func_args@entry=0x0, errstring=errstring@entry=0x555555b74749 "", mask=mask@entry=6) at exceptions.c:557
#10 0x000055555590ad06 in captured_main (data=data@entry=0x7fffffffd730) at main.c:1064
#11 0x00005555559086fa in catch_errors (func=func@entry=0x555555909ed0 <captured_main>, func_args=func_args@entry=0x7fffffffd730, errstring=errstring@entry=0x555555b74749 "", mask=mask@entry=6) at exceptions.c:557
#12 0x000055555590b080 in gdb_main (args=0x7fffffffd730) at main.c:1079
#13 gdb_main_entry (argc=, argv=) at main.c:1099
#14 0x0000555555714d81 in gdb_main_loop (argc=2, argv=0x7fffffffd8c8) at gdb_interface.c:76
#15 0x00005555556568f4 in main (argc=3, argv=0x7fffffffd8c8) at main.c:707

Breakpoint 1, read_mem_section (addr=18446683600716103672) at memory.c:17005
17005 if ((addr == 0) || !IS_KVADDR(addr))
#0 read_mem_section (addr=18446683600716103672) at memory.c:17005
#1 0x00005555556ad30c in valid_section (addr=18446683600716103672) at memory.c:17066
#2 0x00005555556ad4a6 in valid_section_nr (nr=487457) at memory.c:17112
#3 0x00005555556ad654 in dump_mem_sections (initialize=1) at memory.c:17155
#4 0x00005555556aa88c in dump_memory_nodes (initialize=1) at memory.c:16309
#5 0x00005555556ab947 in node_table_init () at memory.c:16536
#6 0x000055555567322c in vm_init () at memory.c:1093
#7 0x0000555555656a98 in main_loop () at main.c:770
#8 0x0000555555909bb3 in captured_command_loop (data=data@entry=0x0) at main.c:258
#9 0x00005555559086fa in catch_errors (func=func@entry=0x555555909ba0 <captured_command_loop>, func_args=func_args@entry=0x0, errstring=errstring@entry=0x555555b74749 "", mask=mask@entry=6) at exceptions.c:557
#10 0x000055555590ad06 in captured_main (data=data@entry=0x7fffffffd730) at main.c:1064
#11 0x00005555559086fa in catch_errors (func=func@entry=0x555555909ed0 <captured_main>, func_args=func_args@entry=0x7fffffffd730, errstring=errstring@entry=0x555555b74749 "", mask=mask@entry=6) at exceptions.c:557
#12 0x000055555590b080 in gdb_main (args=0x7fffffffd730) at main.c:1079
#13 gdb_main_entry (argc=, argv=) at main.c:1099
#14 0x0000555555714d81 in gdb_main_loop (argc=2, argv=0x7fffffffd8c8) at gdb_interface.c:76
#15 0x00005555556568f4 in main (argc=3, argv=0x7fffffffd8c8) at main.c:707
crash: invalid kernel virtual address: ffffc90008b50000 type: "memory section"
[Inferior 1 (process 27504) exited with code 01]
No stack.
(gdb)

[crash-utility]

----- Original Message -----

I can confirm the same issue. The regression appeared with commit d586679. Removing the line "machdep->is_page_ptr = x86_64_is_page_ptr;" after having checked out commit 4141373 does not help, crash fails with the same "memory section" error message.

It's almost impossible for me to debug this, so I'm just guessing at this point. Does the crash-7.2.5 run live on that particular kernel version on that system? Another thing you could check is whether the correct machdep->max_physmem_bits is being set in x86_64.c (line 664). Check whether it matches up with your kernel's MAX_PHYSMEM_BITS setting. If not, there's a --machdep option to override the default setting. BTW, did you check with Petr Tesarik? He does the SUSE-specific maintenance of the crash utility. ([email protected])