kernel hangs after running program that uses a custom SIGSEGV handler
Hi,
$ uname Linux stage4.fedoraproject.org 4.16.0-rc2-00328-gebea62367bc4 #4 SMP Mon Feb 26 16:05:16 GMT 2018 riscv64 riscv64 riscv64 GNU/Linux
This kernel either hangs or prints "BUG" messages when running a program that installs a SIGSEGV handler. That handler uses mprotect to change the access permissions of memory pages. It is known to work fine on all other Linux architectures that I tested.
How to reproduce:
1) Create a VM for execution in QEMU. Details:
http://git.savannah.gnu.org/gitweb/?p=libffcall.git;a=blob;f=porting-tools/emulation/qemu-riscv64.txt
2) Download https://haible.de/bruno/gnu/hangs.tar.xz
and unpack it in the VM, in /.
3) $ cd home/bruno/clisp-2.49.93-20180410/build-porting64-gcc-typecodes-spvw_mixed_blocks-trivialmap
$ ./hangs_linux
Results:
* Either (seen twice) the machine hangs. After 'Ctrl-a x' in qemu and rebooting, you may see files of size 0, due to unsynced open files.
* Or normal output, followed by kernel messages such as
BUG: Bad page map in process
or
BUG: Bad rss-counter stateFull output:
===============================================================================
i i i i i i i ooooo o ooooooo ooooo ooooo
I I I I I I I 8 8 8 8 8 o 8 8
I \ `+' / I 8 8 8 8 8 8
\ `-+-' / 8 8 8 ooooo 8oooo
`-__|__-' 8 8 8 8 8
| 8 o 8 8 o 8 8
------+------ ooooo 8oooooo ooo8ooo ooooo 8Welcome to GNU CLISP 2.49.93+ (2018-02-18) <http://clisp.org/>
Copyright (c) Bruno Haible, Michael Stoll 1992-1993 Copyright (c) Bruno Haible, Marcus Daniels 1994-1997 Copyright (c) Bruno Haible, Pierpaolo Bernardi, Sam Steingold 1998 Copyright (c) Bruno Haible, Sam Steingold 1999-2000 Copyright (c) Sam Steingold, Bruno Haible 2001-2018
Type :h and hit Enter for context help.
;; Loading file ../src/defseq.lisp ...
;; Loaded file ../src/defseq.lisp
;; Loading file ../src/backquote.lisp ...
;; Loaded file ../src/backquote.lisp
;; Loading file ../src/defmacro.lisp ...
;; Loaded file ../src/defmacro.lisp
;; Loading file ../src/macros1.lisp ...
;; Loaded file ../src/macros1.lisp
;; Loading file ../src/macros2.lisp ...
;; Loaded file ../src/macros2.lisp
;; Loading file ../src/defs1.lisp ...
;; Loaded file ../src/defs1.lisp
;; Loading file ../src/lambdalist.lisp ...
;; Loaded file ../src/lambdalist.lisp
;; Loading file ../src/places.lisp ...Unable to handle kernel paging request at virtual address 000000000000c790 Oops [#1]
CPU: 0 PID: 23124 Comm: lisp.run Not tainted 4.16.0-rc2-00328-gebea62367bc4 #4
sepc: ffffffe0005eaec6 ra : ffffffe0005eaf54 sp : ffffffe079a25bd0
gp : ffffffe0007d31e8 tp : ffffffe078f2a300 t0 : ffffffe000034b1a t1 : 0000000000000001 t2 : ffffffe0005f9ea0 s0 : ffffffe079a25be0 s1 : 0000000000000000 a0 : 000000000000c788 a1 : 00000000005414b9 a2 : 0000000000000000 a3 : ffffffe079a25be8 a4 : 00000000005414b9 a5 : ffffffe000883f48 a6 : 000000000000c780 a7 : ffffffe078ea0680 s2 : 0000000000000001 s3 : ffffffffffffffff s4 : 000000000000c788 s5 : 00000000005414b9 s6 : 0000000000000001 s7 : 0000000000000001 s8 : ffffffe0007d4280 s9 : ffffffffffffffff s10: 0000000000000000
s11: ffffffffffffffef t3 : 0000000000000040 t4 : 8000000000000000 t5 : 0000000000000013 t6 : 0000000000000040
sstatus: 0000000000000120 sbadaddr: 000000000000c790 scause: 000000000000000d
---[ end trace 67d406d69d2748af ]--- swap_info_get: Bad swap file entry 2000000000540e29 BUG: Bad page map in process lisp.run pte:2a0714c0 pmd:3e459c01 addr:000000004164e4eb vm_flags:00100070 anon_vma:000000005e8138e7 mapping: (null) index:800e2 file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe swap_info_get: Bad swap file entry 2000000000541129 BUG: Bad page map in process lisp.run pte:2a0894c0 pmd:3e459c01 addr:000000007f6246af vm_flags:00100070 anon_vma:000000005e8138e7 mapping: (null) index:800ec file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G B D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe swap_info_get: Bad swap file entry 20000000005414b9 BUG: Bad page map in process lisp.run pte:2a0a5cc0 pmd:3e459c01 addr:000000007605611e vm_flags:00100073 anon_vma:000000005e8138e7 mapping: (null) index:80101 file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G B D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe swap_info_get: Bad swap file entry 20000000005415d1 BUG: Bad page map in process lisp.run pte:2a0ae8c0 pmd:3e7b2001 addr:00000000e4c69b17 vm_flags:00100070 anon_vma:00000000a7ffaac9 mapping: (null) index:1ffff81 file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G B D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe swap_info_get: Bad swap file entry 2000000000541049 BUG: Bad page map in process lisp.run pte:2a0824c0 pmd:3e7b2001 addr:0000000011fbaac6 vm_flags:00100070 anon_vma:00000000a7ffaac9 mapping: (null) index:1ffffe4 file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G B D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe swap_info_get: Bad swap file entry 2000000000541059 BUG: Bad page map in process lisp.run pte:2a082cc0 pmd:3e7b2001 addr:000000007a53452e vm_flags:00100070 anon_vma:00000000a7ffaac9 mapping: (null) index:1ffffe5 file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G B D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe swap_info_get: Bad swap file entry 20000000005410a9 BUG: Bad page map in process lisp.run pte:2a0854c0 pmd:3e7b2001 addr:00000000e42f76a8 vm_flags:00100070 anon_vma:00000000a7ffaac9 mapping: (null) index:1ffffec file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G B D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe swap_info_get: Bad swap file entry 2000000000540e21 BUG: Bad page map in process lisp.run pte:2a0710c0 pmd:3e7b2001 addr:00000000475e59c6 vm_flags:00100070 anon_vma:00000000a7ffaac9 mapping: (null) index:1fffff8 file: (null) fault: (null) mmap: (null) readpage: (null) CPU: 0 PID: 23124 Comm: lisp.run Tainted: G B D 4.16.0-rc2-00328-gebea62367bc4 #4 Call Trace: [<ffffffe0000341ee>] walk_stackframe+0x0/0xa2 [<ffffffe0000342ec>] show_stack+0x26/0x34 [<ffffffe0005e4904>] dump_stack+0x5e/0x7c [<ffffffe0000c83f4>] print_bad_pte+0x12a/0x192 [<ffffffe0000c9624>] unmap_page_range+0x2ca/0x564 [<ffffffe0000c98e8>] unmap_single_vma+0x2a/0x3e [<ffffffe0000c9a26>] unmap_vmas+0x26/0x46 [<ffffffe0000cf04c>] exit_mmap+0x60/0xf2 [<ffffffe00003567a>] mmput.part.7+0x14/0x84 [<ffffffe000035706>] mmput+0x1c/0x28 [<ffffffe0000398fc>] do_exit+0x184/0x724 [<ffffffe000033e9e>] die+0xd4/0xe4 [<ffffffe000034dc6>] do_page_fault+0x2ac/0x2b8 [<ffffffe000033242>] ret_from_syscall+0xa/0xe BUG: Bad rss-counter state mm:00000000c900d1f0 idx:1 val:8 BUG: Bad rss-counter state mm:00000000c900d1f0 idx:2 val:-8
Segmentation fault
Quoted 9 lines · Bruno Haible
Hi,
$ uname Linux stage4.fedoraproject.org 4.16.0-rc2-00328-gebea62367bc4 #4 SMP Mon Feb 26 16:05:16 GMT 2018 riscv64 riscv64 riscv64 GNU/LinuxThis kernel either hangs or prints "BUG" messages when running a program that installs a SIGSEGV handler. That handler uses mprotect to change the access permissions of memory pages. It is known to work fine on all other Linux architectures that I tested.
The bug is caused by mishandling of PROT_NONE (failing to distinguish between PROT_NONE and non-present pages), not by the segfault handler. I sent a patch for this to Andrew and Richard a few weeks ago but haven't had a chance to follow up.
-s
Quoted 2 lines · Stef O'Rear
The bug is caused by mishandling of
PROT_NONE(failing to distinguish betweenPROT_NONEand non-present pages)
Indeed, this sounds like a possible bug cause: clisp is one of the few programs that uses mprotect not only with PROT_READ and PROT_READ_WRITE but also occasionally with PROT_NONE.
Bruno
Alexandre Ghiti
reviewerHi Stef,
I'm happy to see there are solutions to this problem: I have a working port of hugetlb for riscv-linux and I was wondering how to support PROT_NONE.
Can you share a link to the patches you mention please ?
Thanks,
Alex
Quoted 21 lines · Stef O'Rear
Quoted 9 lines · Bruno Haible
Hi,
$ uname Linux stage4.fedoraproject.org 4.16.0-rc2-00328-gebea62367bc4 #4 SMP Mon Feb 26 16:05:16 GMT 2018 riscv64 riscv64 riscv64 GNU/LinuxThis kernel either hangs or prints "BUG" messages when running a program that installs a SIGSEGV handler. That handler uses mprotect to change the access permissions of memory pages. It is known to work fine on all other Linux architectures that I tested.
The bug is caused by mishandling of
PROT_NONE(failing to distinguish betweenPROT_NONEand non-present pages), not by the segfault handler. I sent a patch for this to Andrew and Richard a few weeks ago but haven't had a chance to follow up.-s
_______________________________________________
linux-riscv mailing list
linux-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
Alexandre Ghiti <[email protected]> ? 2018?4?13? ?? ??6:14???
Quoted 7 lines
Hi Stef,
I'm happy to see there are solutions to this problem: I have a working port of hugetlb for riscv-linux and I was wondering how to support
PROT_NONE.
Can you share a link to the patches you mention please ?
Quoted 11 lines
Thanks,
Alex
Quoted 5 lines · Stef O'Rear
Quoted 4 lines · Bruno Haible
Hi,
$ uname Linux stage4.fedoraproject.org 4.16.0-rc2-00328-gebea62367bc4 #4 SMP
Mon Feb 26 16:05:16 GMT 2018 riscv64 riscv64 riscv64 GNU/Linux
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
This kernel either hangs or prints "BUG" messages when running a
program that
Quoted 1 line
Quoted 1 line
Quoted 1 line
installs a SIGSEGV handler. That handler uses mprotect to change the
access
Quoted 1 line
Quoted 1 line
Quoted 1 line
permissions of memory pages. It is known to work fine on all other
Linux
Quoted 14 lines
Quoted 6 lines
Quoted 1 line
architectures that I tested.
The bug is caused by mishandling of
PROT_NONE(failing to distinguish betweenPROT_NONEand non-present pages), not by the segfault handler. I sent a patch for this to Andrew and Richard a few weeks ago but haven't had a chance to follow up.Quoted 6 lines
-s
_______________________________________________
linux-riscv mailing list
linux-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
Quoted 4 lines
_______________________________________________
linux-riscv mailing list
linux-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
Is there any update? I also encounter this issue, and I change the permission like below.
Where is the patch as Stef O'Rear mention before?
(quoted diff) +0 -0
b/arch/riscv/include/asm/pgtable.h
index 16301966d65b..578f31510501 100644
arch/riscv/include/asm/pgtable.h +1 -1
@@ -44,7 +44,7 @@/* Page protection bits */ #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_USER) -#define PAGE_NONE __pgprot(0) +#define PAGE_NONE __pgprot(_PAGE_BASE)