I am working on an embedded Linux system (kernel-5.10.24), the CPU is 32bit MIPs.
The applications run in the system may trigger invalid memory access, which will be shot by a SIGSEGV from kernel, and the kernel may dump some logs as follows,
[ 5.464129] do_page_fault(): sending SIGSEGV to testsegv for invalid read access from 0000001c
[ 5.464144] epc = 0041e118 in testsegv[400000+668000]
[ 5.464173] ra = 00661010 in testsegv[400000+668000]
This log is too simple to triage the problem.
So I am trying to do some backtracing in kernel by adding show_regs() into the mm/fault.c and I can got following logs when hit error.
[ 185.408332] do_page_fault(): sending SIGSEGV to segv for invalid write access to 00000000
[ 185.418592] epc = 0040065c in segv[400000+1000]
[ 185.423642] ra = 00400654 in segv[400000+1000]
[ 185.428349] CPU: 1 PID: 1235 Comm: segv Not tainted 5.10.24 #17
[ 185.434760] $ 0 : 00000000 00000001 00000000 00000063
[ 185.440325] $ 4 : 77e7953c 00c8a190 ffffffff 00000000
[ 185.445742] $ 8 : 00000000 00000000 00000001 68736172
[ 185.451338] $12 : 0000000d 00000080 00000000 00000000
[ 185.456755] $16 : 7faf6564 77ecaf10 004007a0 00000002
[ 185.462340] $20 : 00000000 77ec6508 77ecb408 00400714
[ 185.467748] $24 : 00000000 77d38a60
[ 185.473200] $28 : 77e7ee30 7faf63d0 7faf63d0 00400654
[ 185.478712] Hi : 00000013
[ 185.481717] Lo : 00000000
[ 185.484774] epc : 0040065c 0x40065c
[ 185.488555] ra : 00400654 0x400654
[ 185.492383] Status: 04001c13 USER EXL IE
[ 185.496639] Cause : 0880000c (ExcCode 03)
[ 185.500797] BadVA : 00000000
[ 185.508620] CPU: 1 PID: 1235 Comm: segv Not tainted 5.10.24 #17
[ 185.514826] Stack : 80bb0000 80092358 00000000 00000000 80a8ee08 8160f950 81deb528 80095370
[ 185.523482] 00000000 00000000 00000000 7b699e72 825dbe6c 00000001 825dbe00 7b699e72
[ 185.532132] 00000000 00000000 80a74f30 825dbcc0 000001cf 825dbcd4 00000000 00001388
[ 185.540785] 1e50ef51 825dbcd3 ffffffff 00000030 80b90000 80000000 00000000 80a70000
[ 185.549432] 00000001 00000001 8160f900 8160f950 00000000 00000000 2000e098 80c40004
[ 185.558081] ...
[ 185.560613] Call Trace:
[ 185.563149] [<8001f294>] show_stack+0x94/0x12c
[ 185.567745] [<8094857c>] dump_stack+0xac/0xe8
[ 185.572253] [<8002c714>] do_page_fault+0x2d4/0x510
[ 185.577211] [<80032b98>] tlb_do_page_fault_1+0x118/0x120
It showed the backtrace in kernel space not the user space.
So is there a way to get the backtrace of user space in Linux kernel in this case? (IIRC, X86 can dump something more in kernel space).