mbox series

[RFC,0/7] linux-user: Streamline handling of SIGSEGV/SIGBUS

Message ID 20210913220552.604064-1-richard.henderson@linaro.org
Headers show
Series linux-user: Streamline handling of SIGSEGV/SIGBUS | expand

Message

Richard Henderson Sept. 13, 2021, 10:05 p.m. UTC 
Our current setup is:

  host_signal_handler
  cpu_signal_handler
     handle_cpu_signal
        cc->tcg_ops->tlb_fill
           raise_exception
  cpu_loop
     queue_signal

and in the process lose information from the host siginfo_t,
which we (mostly) do not recreate properly.  Moreover, the
intermediate cpu_signal_handler handles the host-specific
portions of extracting pc + is_write from the ucontext_t.

I'm replacing this with

  host_signal_handler
    host_signal_pc
    host_sigsegv_write
    adjust_signal_pc
    handle_sigsegv_accerr_write
    queue_signal
    raise_exception
  cpu_loop

All of the really tcg-specific portions are still in user-exec.c,
and all of the really host-specific portions are now ditributed
across linux-user/host/<arch>/.  Importantly, SEGV_MAPERR and
SEGV_ACCERR are now passed through from the host kernel -- or at
least there's a single place from which to manage it [1].

Note that I've dropped all of the BSD (and Solaris!) code from
user-exec.c.  I thought about moving it similar to linux-user,
but I've caught Warner in the middle of his re-org and the whole
of bsd-user/signal.c is currently empty.  I think it will be
easier to create the new interfaces from scratch when ready.

Still to-do:
  * Make cc->tcg_ops->tlb_fill sysemu only (once again).
  * Drop all of the code from cpu_loop that interfaced with tlb_fill.


r~


[1] I've just this minute realized that the reserved_va mapping that we
do for emulating 32-bit guests will incorrectly give SEGV_ACCERR for pages
that are not mapped by the guest, and should result in SEGV_MAPERR.


Richard Henderson (7):
  include/exec: Move cpu_signal_handler declaration
  accel/tcg: Split out adjust_signal_pc
  accel/tcg: Split out handle_sigsegv_accerr_write
  accel/tcg: Move clear_helper_retaddr to cpu loop
  accel/tcg: Fold cpu_exit_tb_from_sighandler into caller
  linux-user: Handle SIGSEGV/SIGBUS in host_to_target_siginfo_noswap
  linux-user: Reorg cpu_signal_handler

 include/exec/exec-all.h               |  21 +
 linux-user/host/aarch64/host-signal.h |  73 +++
 linux-user/host/alpha/host-signal.h   |  41 ++
 linux-user/host/arm/host-signal.h     |  30 +
 linux-user/host/i386/host-signal.h    |  24 +
 linux-user/host/mips/host-signal.h    |  61 ++
 linux-user/host/ppc/host-signal.h     |  24 +
 linux-user/host/ppc64/host-signal.h   |   1 +
 linux-user/host/riscv32/host-signal.h |  57 ++
 linux-user/host/riscv64/host-signal.h |   1 +
 linux-user/host/s390/host-signal.h    |  92 +++
 linux-user/host/s390x/host-signal.h   |   1 +
 linux-user/host/sparc/host-signal.h   |  53 ++
 linux-user/host/sparc64/host-signal.h |   1 +
 linux-user/host/x86_64/host-signal.h  |  24 +
 target/alpha/cpu.h                    |   6 -
 target/arm/cpu.h                      |   7 -
 target/avr/cpu.h                      |   2 -
 target/cris/cpu.h                     |   8 -
 target/hexagon/cpu.h                  |   3 -
 target/hppa/cpu.h                     |   3 -
 target/i386/cpu.h                     |   7 -
 target/m68k/cpu.h                     |   8 -
 target/microblaze/cpu.h               |   7 -
 target/mips/cpu.h                     |   3 -
 target/mips/internal.h                |   2 -
 target/nios2/cpu.h                    |   2 -
 target/openrisc/cpu.h                 |   2 -
 target/ppc/cpu.h                      |   7 -
 target/riscv/cpu.h                    |   2 -
 target/rx/cpu.h                       |   4 -
 target/s390x/cpu.h                    |   7 -
 target/sh4/cpu.h                      |   3 -
 target/sparc/cpu.h                    |   2 -
 target/tricore/cpu.h                  |   2 -
 target/xtensa/cpu.h                   |   2 -
 accel/tcg/cpu-exec.c                  |   3 +-
 accel/tcg/user-exec.c                 | 807 ++------------------------
 linux-user/signal.c                   | 102 ++--
 39 files changed, 635 insertions(+), 870 deletions(-)
 create mode 100644 linux-user/host/aarch64/host-signal.h
 create mode 100644 linux-user/host/alpha/host-signal.h
 create mode 100644 linux-user/host/arm/host-signal.h
 create mode 100644 linux-user/host/i386/host-signal.h
 create mode 100644 linux-user/host/mips/host-signal.h
 create mode 100644 linux-user/host/ppc/host-signal.h
 create mode 100644 linux-user/host/ppc64/host-signal.h
 create mode 100644 linux-user/host/riscv32/host-signal.h
 create mode 100644 linux-user/host/riscv64/host-signal.h
 create mode 100644 linux-user/host/s390/host-signal.h
 create mode 100644 linux-user/host/s390x/host-signal.h
 create mode 100644 linux-user/host/sparc/host-signal.h
 create mode 100644 linux-user/host/sparc64/host-signal.h
 create mode 100644 linux-user/host/x86_64/host-signal.h

-- 
2.25.1

Comments

Richard Henderson Sept. 14, 2021, 1:18 a.m. UTC | #1 
On 9/13/21 3:05 PM, Richard Henderson wrote:
> All of the really tcg-specific portions are still in user-exec.c,

> and all of the really host-specific portions are now ditributed

> across linux-user/host/<arch>/.  Importantly, SEGV_MAPERR and

> SEGV_ACCERR are now passed through from the host kernel -- or at

> least there's a single place from which to manage it [1].


Hum.  And then there's the special case of s390x, where the hw does not provide the exact 
address on faults, but only the page.  We have code for that in cpu_loop, but I have to 
invent some new hook in this new scheme.


r~