Re: Possible reasons for SIGBUS error?

MWimmer <michael.wimmer1@xxxxxx> writes:

[...]

Since then, each of us has independently experienced crashes with
signal SIGBUS. These crashes were not reproducible: Running the same
program again did not result in an error. The machines used were
Pentium 4's and Xeons (We have yet to make a statistic, if the error
only happens on certain machines)

I've been searching on the internet for possible causes of SIGBUS. It
seems, that there are two possible occasions:

1. misaligned memory pointer. However, according to what I read on the
internet, the Pentium should also be able to cope with misaligned
memory.
2. Something wrong with respect to mmap(). I'm not an expert in this,
so maybe someone could expolain me that connection a bit more.
However, none of our programs is using mmap() directly.

The only occurences of SIGBUS on x86-Linux I had so far where caused
by bad memory. You definitely shouldn't get alignment traps and
without some amount of criminal energy[1], compiled code shouldn't do
misaligned accesses, anyway.

[1] example:

char a[8];
unsigned *p;

p = (unsigned *)(a + 1);
*p = 0

That would likely trap an architecture that can't do misaligned
accesses.

.

Relevant Pages

  • Re: Toshiba Satellite 320 specs
    ... I am extremely familiar with this model (Toshiba Satellite 320 series), ... All models also have an SO-DIMM slot for additonal memory under the keyboard. ... These are truly great older machines. ... It was superseeded directly by the early 4000 series, which are actually the exact same machines but with Pentium II's and SDRAM memory instead of Pentium MMX's and EDO memory. ...
    (comp.sys.laptops)
  • SVCHOST memory leak?
    ... After doing updates from June and July, a few of my machines in the ... following configuration are throwing up messages about low virtual memory. ... Pentium 4 2.0GHz machine ... The following are the svchost processes, ...
    (microsoft.public.windowsxp.general)
  • Re: Possible reasons for SIGBUS error?
    ... signal SIGBUS. ... Pentium 4's and Xeons (We have yet to make a statistic, ... only happens on certain machines) ... by bad memory. ...
    (comp.os.linux.development.apps)
  • Re: Handling disk I/O errors
    ... > they would get a SIGBUS reading the same part of the file every single time. ... You don't get SIGBUS from reading a file. ... > Yes, memory is often bad, but this doesn't seem to be a memory problem. ... > whereabouts which block was bad on the disk. ...
    (comp.os.linux.development.apps)
  • Re: Possible reasons for SIGBUS error?
    ... We are a group of Linux users that do numerical simulations, ... long and memory intensive compute jobs. ... I've been searching on the internet for possible causes of SIGBUS. ... change the values of parameters passed to a subroutine ...
    (comp.os.linux.development.apps)