CVS log for src/sys/arch/usermode/usermode/trap.c

Up to [cvs.NetBSD.org] / src / sys / arch / usermode / usermode

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

fix various typos in comments, mainly origional->original,
extened->extended, incomming->incoming.

fix various typos, mainly in comments.

Merge changes from current as of 20200406

PR port-sparc/54718 (sparc install hangs since recent scheduler changes)

- userret() must be called every time we return to user, it's not optional.
- If clearing the AST with interrupts off, you must loop over userret().

mi_userret(): take care of calling preempt(), set spc_curpriority directly,
and remove MD code that does the same.

Sync with HEAD

Sync with HEAD

Resolve a couple of conflicts (result of the uimin/uimax changes)

Oops, forgot a debug printf

Fix typo

Sync with HEAD

Prevent nested SIGIOs ruining the system stack

Sync with HEAD

Replace the dynamically allocated signal stack to a static claimed one and add
diagnostic messages that can be enabled.

sync with head

Pull up following revision(s) (requested by reinoud in ticket #463):
	sys/arch/usermode/usermode/trap.c: revision 1.66
Fix IO lockups in NetBSD/usermode.
1) Don't block IO signals since the return path is not garanteed to enable the
signal again.
2) Since signals can get dropped, do a 2nd pass over the routines.

Fix IO lockups in NetBSD/usermode.

1) Don't block IO signals since the return path is not garanteed to enable the
signal again.
2) Since signals can get dropped, do a 2nd pass over the routines.

sync with head

Pull up following revision(s) (requested by reinoud in ticket #91):
	sys/arch/usermode/usermode/trap.c: revision 1.65
Use the signal's signo instead of just returning SIGSEGV and add a comment

Pull up following revision(s) (requested by reinoud in ticket #89):
	sys/arch/usermode/usermode/thunk.c: revision 1.81
	sys/arch/usermode/target/i386/cpu_i386.c: revision 1.4
	sys/arch/usermode/include/thunk.h: revision 1.61
	sys/arch/usermode/include/intr.h: revision 1.9
	sys/arch/usermode/dev/cpu.c: revision 1.70
	sys/arch/usermode/dev/ttycons.c: revision 1.19
	sys/arch/usermode/usermode/trap.c: revision 1.64
Implement signal forwarding to userland for usermode kernels. Especially
SIGFPE is important since the urkel shouldn't stop when a userland program
devides by zero!

sync to -current

sync to -current

sync to latest -current.

Use the signal's signo instead of just returning SIGSEGV and add a comment

Implement signal forwarding to userland for usermode kernels. Especially
SIGFPE is important since the urkel shouldn't stop when a userland program
devides by zero!

merge to -current.

Prevent recursive IO signals. This prevents crashes when pieces of text are
pasted in say an Xterm. Each char might create a SIGIO running out of
stackspace very quickly!

Fix compilation issue for target amd64.

Complete rewrite of the signal and spl framework for NetBSD/usermode

Signals are now moved from the sigaltstack ASAP and stacked on a replacement
stack for each processes.

Preemption now works though could be enhanced a bit more

Another step towards unification of the signal handers: separate the printing
of details usefull for debugging / porting from the main code.

Do init the astpending variable.

Clean up userret/ast handling. Now called ONLY when its 100% clear that it is
returning to userland.

Reimplement userret() to also include AST for preemption.

Note it would be nice if we could do the check "are we going to userland?" in
a less intrusive way.

Revamp the NetBSD/usermode pagefault and illegal instruction handing. It now
can handle recursive entry and is a lot more memory tight compared to the old
implementation. Performance wise:

* slightly less number of syscalls/sec possible though could be optimized
* a lot faster context creation / destruction making overall operation faster.

Cleanup trap.c and add a stop_all_signal_handlers(void) i: bluntly stopping
all known usages of signal handlers, to be used in shutdown.

Cleanup trap debug messages

printf()->thunk_printf() since its for debugging and should be printed outside
the kernel's printing system.

Rework NetBSD/usermode pmap fixing some oddities that were left over from
earlier times when we were forced to run PIE executables and were forced to
use a KVM above the kernel.

Rename the debug printf's to use a thunk_ prefix to avoid confusion.

Fix logic that checks if its the kernel or a userland process that causes the
trap. Signal sending still needs to be addressed.

Remove tests that are dubious and should be (re)moved anyway.

make sure the sigio signal handler runs on the alternate signal stack,
fixes random SIGILLs seen recently

add sigio_intr_establish so more than one driver can register a SIGIO handler

Fix VM_MAX_ADDRESS that is NOT the max address BUT the max userland address
Fix VM_MAX_KERNEL_ADDRESS that is the absolute max address accesable by the
kernel...

Remove the need for a siginfo structure in illegal instruction handling.

need to sigaltstack for each lwp

Remove old debugging cruft

Big patch that changes the signal stack usage of urkel significantly.

Formerly, all signals came on the signal stack and the two important ones were
then forwared to either the system call or the pagefault handler. This worked
fine but the signal stack remains that, a stack. When we go multi-process this
stack gets corrupted and out-of-order with all kind of nastyness since a
userland process switch can occure when a system call is called or when a
process gets a page fault.

The new scheme only uses the signal stack as a jumpboard. It swaps states and
then returns from the signal, clearing the stack but instead of returning to
the code it now jumpt to the handler and that handler then returns to the code
when its finished.

Move the debug printf's to the new dprintf_debug() and disable the SIGALRM
blocking since we queue them now.

Move to dprintf_debug()

The illegal instruction handler can exist on the usermodes stack and the
stackframe will disappear when its switched back.

Be more verbose in trap.c on what address should be passed to userland. Also
detail about what function goes into recurse

Implement a working(!) syscall! well, it can call a number of syscalls without
dying ;)

Use the machdep.h and now use the md_check_syscall_opcode()

Create syscall() prototype and let illegal instruction handler switch to that
switchframe

Remove struct trapframe

Add skeleton for possible system call implementation

Better not copy this trapframe

Print the byte codes that failed

State actions to implement.

cpu_lwp_inkernel approach won't work, back to the drawing board

Move sanity check before pmap since pmap and uvm can't be looking outside this
box

adjust cpu_lwp_inkernel in mem access and illegal insn handlers

Check for NULL pcb!

Use thunk's version of sigemptyset()

Initialise the alternative signal stack before referring to it

Block SIGALRM on memory access and illegal instruction signal

pcb->errno to pcb->pcb_errno

Try to preserve errno over traps and on context switches

mem_access_handler: panic if we receive sigsegv or sigbus with SI_NOINFO,
and when "peeing outside the box" print the offending va

Add SIGILL handler, doing nothing than printing out details.
Move the address check to be done later so pmap and uvm have their go first.

Also some minor cosmetic fixes

Get this compiling (but not quite linking yet) on Linux

Add note on how to posibly create a signal and send it to a userland program

Disable the debug_fh code for now

Use a alternative dedicated signal stack for the sigsegv errors; the userland
stack like that of ld.elf_so can be too short to allow for such deep
stacktraces.

Enhance debugging capabilities of NetBSD/usermode by allowing `trap.c' to
write a system image of the memory it traps. Its extremely dirty but at least
you can look at the data and code thats giving the trouble. Its memory or
state is NOT saved in the coredump of the NetBSD/usermode program.

Completely redo R/M emulation and fault handling taking UVM as authorative
answer and not relying on UVM's pmap directions since UVM has its own schemes
for COW etc.

Implement read and modify detection code

mem_access_handler: print originating lwp ptr + pid + lid

Hack trap for now to upgrade protection.. this really needs fixing soon

Don't use the content of trapframe yet until we know how to use fill it

cpu_lwp_trampoline isn't supposed to return, so drop the atexit handler
and drop a panic() at the end of the function

Catch the atexit() so we can bypass the libc feature that if one ucontext()
ends that it then call exit()

Implement trap recursion detection and fix small one-off error in range
checks for kernel space

A trap is now allways asked for RW access... that needs to be fixed but thats
later care.

Check if we pee'd outside the box!

Update and claim copyright

Implement first generation of memory access handler for the kernel.

sync with head

Drop 3rd and 4th clauses.  OK jmcneill (copyright holder).

Sync with HEAD.

sync with head

sync with HEAD

Sync with HEAD

file trap.c was added on branch mjf-devfs on 2008-02-18 21:05:13 +0000

file trap.c was added on branch yamt-lazymbuf on 2008-01-21 09:39:58 +0000

file trap.c was added on branch matt-armv6 on 2008-01-09 01:49:31 +0000

file trap.c was added on branch bouyer-xeni386 on 2008-01-02 21:51:07 +0000

Import work-in-progress NetBSD/usermode port.