Date: Mon, 1 Apr 1996 13:40:34 -0500 (EST) From: Andrew Gallatin <gallatin@isds.Duke.EDU> To: Colman Reilly <creilly@maths.tcd.ie> Cc: hackers@freebsd.org Subject: Re: Linux emulator and Mathematica Message-ID: <199604011840.NAA19545@davinci.isds.duke.edu> In-Reply-To: <9604011003.aa25264@salmon.maths.tcd.ie> References: <9604011003.aa25264@salmon.maths.tcd.ie>
next in thread | previous in thread | raw e-mail | index | archive | help
Colman Reilly writes:
> The network. To make the network run, you have to emulate the SIOCHWADDR call
> from linux. This returns the ethernet address of your machine.
>
> Um. Horrible code segment follows.
>
> Insert into linux_ioctl, in linux_ioctl.c, inside the first switch.
>
> case 0x8927:
> hwaddr[0]=0x00;
> hwaddr[1]=0x00;
> hwaddr[2]=0xc0;
> hwaddr[3]=0x27;
> hwaddr[4]=0xe5;
> hwaddr[5]=0x66;
> return copyout((caddr_t)hwaddr, (caddr_t)args->arg,6);
>
>
> The 6 numbers need to be your ethernet address in hex. I'm sorry, this is a
> disgusting hack, but I still haven't had time to fix it
>
> Colman
Thanks... Actually, I consider being able to hardcode the HW address
a feature ;-) We have a university wide site-license for mathematica,
but I still hate having to deal with the key givers..
Anyway, I've slapped the above into linux_ioctl, in linux_ioctl.c (btw,
what type is hwaddr supposed to be?, I've got it declared as int
hwaddr[6]) I've appended what I've put into the 2.1R linux_dummy.c.
Its taken nearly verbatum from the the 3/23 snap.
At this point, I've attemped to use both a network & standalone
license. I'd really rather get the standalone license working..
The network license sort of works, but strange goblety-gook
is printed to the screen, it looses connections to the server, and
mathclients are spawned which go into a cpu-bound infinate loop.
The standalone (which is what I really want) stops on the error:
General::codespace: Code space corrupted
a ktrace shows it reading from a file
(/usr/local/math/Install/Preload/msgthin.m), closing it then dying:
213 mathexe RET read 2833/0xb11
213 mathexe CALL close(3)
213 mathexe RET close 0
213 mathexe CALL write(1,4149248,1)
213 mathexe GIO fd 1 wrote 1 bytes
"
"
213 mathexe RET write 1
213 mathexe CALL write(1,4149248,41)
213 mathexe GIO fd 1 wrote 41 bytes
"General::codespace: Code space corrupted."
213 mathexe RET write 41/0x29
213 mathexe CALL write(1,4149248,1)
213 mathexe GIO fd 1 wrote 1 bytes
"
"
213 mathexe RET write 1
213 mathexe PSIG SIGBUS SIG_DFL
Where do I go from here?
Thanks,
Drew
##############################################################################
# Andrew Gallatin, Computer Project Manager #
# Institute of Statistics and Decision Sciences #
# Box 90251, Duke University, Durham, NC 27708-0251 #
##############################################################################
#include <sys/signal.h>
#include <sys/sysctl.h>
#include <sys/proc.h>
#include <sys/user.h>
#include <machine/proc.h>
#include <machine/cpu.h>
#include <machine/reg.h>
#include <machine/psl.h>
struct linux_sigcontext {
int sc_gs;
int sc_fs;
int sc_es;
int sc_ds;
int sc_edi;
int sc_esi;
int sc_ebp;
int sc_esp;
int sc_ebx;
int sc_edx;
int sc_ecx;
int sc_eax;
int sc_trapno;
int sc_err;
int sc_eip;
int sc_cs;
int sc_eflags;
int sc_esp_at_signal;
int sc_ss;
int sc_387;
int sc_mask;
int sc_cr2;
};
struct linux_sigreturn_args {
struct linux_sigcontext * scp;
};
/*
* System call to cleanup state after a signal
* has been taken. Reset signal mask and
* stack state from context left by sendsig (above).
* Return to previous pc and psl as specified by
* context left by sendsig. Check carefully to
* make sure that the user has not modified the
* psl to gain improper privileges or to cause
* a machine fault.
*/
int
linux_sigreturn(struct proc *p, struct linux_sigreturn_args *args, int *retval)
/* struct proc *p;
struct linux_sigreturn_args *args;
int *retval;
*/
{
struct linux_sigcontext *scp, context;
register int *regs;
int eflags;
regs = p->p_md.md_regs;
#ifdef DEBUG
printf("Linux-emul(%d): linux_sigreturn(%8x)\n", p->p_pid, args->scp);
#endif
/*
* The trampoline code hands us the context.
* It is unsafe to keep track of it ourselves, in the event that a
* program jumps out of a signal handler.
*/
scp = args->scp;
if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0)
return (EFAULT);
/*
* Check for security violations.
*/
#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
eflags = context.sc_eflags;
/*
* XXX do allow users to change the privileged flag PSL_RF. The
* cpu sets PSL_RF in tf_eflags for faults. Debuggers should
* sometimes set it there too. tf_eflags is kept in the signal
* context during signal handling and there is no other place
* to remember it, so the PSL_RF bit may be corrupted by the
* signal handler without us knowing. Corruption of the PSL_RF
* bit at worst causes one more or one less debugger trap, so
* allowing it is fairly harmless.
*/
if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs[tEFLAGS] & ~PSL_RF)) {
return(EINVAL);
}
/*
* Don't allow users to load a valid privileged %cs. Let the
* hardware check for invalid selectors, excess privilege in
* other selectors, invalid %eip's and invalid %esp's.
*/
#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
if (!CS_SECURE(context.sc_cs)) {
trapsignal(p, SIGBUS, T_PROTFLT);
return(EINVAL);
}
p->p_sigacts->ps_sigstk.ss_flags &= ~SA_ONSTACK;
p->p_sigmask = context.sc_mask &~
(sigmask(SIGKILL)|sigmask(SIGCONT)|sigmask(SIGSTOP));
/*
* Restore signal context.
*/
/* %fs and %gs were restored by the trampoline. */
regs[tES] = context.sc_es;
regs[tDS] = context.sc_ds;
regs[tEDI] = context.sc_edi;
regs[tESI] = context.sc_esi;
regs[tEBP] = context.sc_ebp;
regs[tEBX] = context.sc_ebx;
regs[tEDX] = context.sc_edx;
regs[tECX] = context.sc_ecx;
regs[tEAX] = context.sc_eax;
regs[tEIP] = context.sc_eip;
regs[tCS] = context.sc_cs;
regs[tEFLAGS] = eflags;
regs[tESP] = context.sc_esp_at_signal;
regs[tSS] = context.sc_ss;
return (EJUSTRETURN);
}
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604011840.NAA19545>