Disabling interrupts

Disabling interrupts

[Pavel Roskin]

Hello!

I need to disable interrupts for a short time from a user program on
RPX/Lite. This needs to be done only once, but it's absolutely essential
that the program is not interrupted during a certain period of time.

Using the __cli() function from the kernel sources in a user-space program
doesn't work. The first instruction "mfmsr %r0" (Get current interrupt
state) causes SIGILL.

Probably I could hack the kernel to make processes with very high priority
uninterruptible.

I'm quite sure that the problem is not new. Anybody has a better idea?

Pavel Roskin


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Disabling interrupts

[Cort Dougan]

Does it have to be a user mode piece of code?  You're going to have real
trouble figuring that out if so.  Could you put that small portion of code
into a kernel module and just load it from your user program to do what it
needs to?

What is it you're trying to do that shouldn't be interrupted?

} I need to disable interrupts for a short time from a user program on
} RPX/Lite. This needs to be done only once, but it's absolutely essential
} that the program is not interrupted during a certain period of time.
}
} Using the __cli() function from the kernel sources in a user-space program
} doesn't work. The first instruction "mfmsr %r0" (Get current interrupt
} state) causes SIGILL.
}
} Probably I could hack the kernel to make processes with very high priority
} uninterruptible.
}
} I'm quite sure that the problem is not new. Anybody has a better idea?
}
} Pavel Roskin
}
}

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Disabling interrupts

[Pavel Roskin]

Hello, Cort!

> Does it have to be a user mode piece of code?  You're going to have real
> trouble figuring that out if so.  Could you put that small portion of code
> into a kernel module and just load it from your user program to do what it
> needs to?

Well, this code needs to read some data continously. Then the data must
be processed and written to a file.

If you expect problems, I could create a new syscall and pass a buffer to
it. Probably this should be Ok. But the buffer can be huge (comparable
with the RAM size), so I would like to avoid copying data. This means that
the kernel needs to write to the userspace buffer directly.

Thank you for the reply.

Regards,
Pavel Roskin


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Disabling interrupts

[Greg Johnson]

Why don't you just use DMA? Have a couple of buffers and flip between
them writing the data to disk while the DMA controller handles the
reads. Again, this will mean writing a module.

Greg.

Quoth Pavel Roskin:
> > Does it have to be a user mode piece of code?  You're going to have real
> > trouble figuring that out if so.  Could you put that small portion of code
> > into a kernel module and just load it from your user program to do what it
> > needs to?
>
> Well, this code needs to read some data continously. Then the data must
> be processed and written to a file.
>
> If you expect problems, I could create a new syscall and pass a buffer to
> it. Probably this should be Ok. But the buffer can be huge (comparable
> with the RAM size), so I would like to avoid copying data. This means that
> the kernel needs to write to the userspace buffer directly.

--
+----------------------------------------------------+
| Do you want to know more?                          |
|               --== Greg Johnson ==--               |
| HW/SW Engineer      gjohnson@research.canon.com.au |
| Canon   Information  Systems   Research  Australia |
| 1 Thomas Holt Dr, North Ryde, NSW, 2113, Australia |
| "I FLEXed my BISON and it went YACC!" - me.        |
+----------------------------------------------------+

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Disabling interrupts

[Cort Dougan]

For something like this rtl may be better suited for the task.

} Well, this code needs to read some data continously. Then the data must
} be processed and written to a file.
}
} If you expect problems, I could create a new syscall and pass a buffer to
} it. Probably this should be Ok. But the buffer can be huge (comparable
} with the RAM size), so I would like to avoid copying data. This means that
} the kernel needs to write to the userspace buffer directly.
}
} Thank you for the reply.
}
} Regards,
} Pavel Roskin
}
}

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/