Re: nswap
Thus spake Mark Santcroos <[EMAIL PROTECTED]>:
> On Tue, Jan 21, 2003 at 01:33:01AM -0800, David Schultz wrote:
> > Thus spake Mark Santcroos <[EMAIL PROTECTED]>:
> > > Yes, that is also what I meant. We now have a swapoff() system call that
> > > does all the work itself.
> > >
> > > My idea was to split that up:
> > >
> > > /* turn of swap device */
> > > static int swapoff_one(struct swdevt *sp)
> > > {
> > > /* Do all things that we don't want to know about outside this function
> > >*/
> > > }
> > >
> > > /* turn off all swap devices */
> > > int swapoff_all()
> > > {
> > > int index;
> > > struct swdevt *sp;
> > >
> > > for (sp = swdevt, index = 0; index < nswdev; index++, sp++)
> > > swapoff_one(sp);
> > > }
> > >
> > > So the swapoff() system call would call swapoff_one() and my code in
> > > kern/kern_swsuspend.c would call swapoff_all().
> >
> > See swapoff(8), in particular the -a flag.
>
> I'm aware of that, but imho it doesn't suite my purpose. Are you strongly
> against having such code inside the kernel?
In that case, I'm not sure I understand what you're trying to do.
There's already a sysctl interface to get the names of all the
swap devices in the system, and from there you can call swapoff(2)
on each one. Anyway, it doesn't really bother me to add similar
functionality to the kernel if that seems useful, but it's also
not my decision. Perhaps if you posted patches, someone could
make more specific comments.
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
Re: nswap
On Tue, Jan 21, 2003 at 01:33:01AM -0800, David Schultz wrote:
> Thus spake Mark Santcroos <[EMAIL PROTECTED]>:
> > Yes, that is also what I meant. We now have a swapoff() system call that
> > does all the work itself.
> >
> > My idea was to split that up:
> >
> > /* turn of swap device */
> > static int swapoff_one(struct swdevt *sp)
> > {
> > /* Do all things that we don't want to know about outside this function
> > */
> > }
> >
> > /* turn off all swap devices */
> > int swapoff_all()
> > {
> > int index;
> > struct swdevt *sp;
> >
> > for (sp = swdevt, index = 0; index < nswdev; index++, sp++)
> > swapoff_one(sp);
> > }
> >
> > So the swapoff() system call would call swapoff_one() and my code in
> > kern/kern_swsuspend.c would call swapoff_all().
>
> See swapoff(8), in particular the -a flag.
I'm aware of that, but imho it doesn't suite my purpose. Are you strongly
against having such code inside the kernel?
Mark
--
Mark SantcroosRIPE Network Coordination Centre
http://www.ripe.net/home/mark/New Projects Group/TTM
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
Re: nswap
Thus spake Mark Santcroos <[EMAIL PROTECTED]>:
> Yes, that is also what I meant. We now have a swapoff() system call that
> does all the work itself.
>
> My idea was to split that up:
>
> /* turn of swap device */
> static int swapoff_one(struct swdevt *sp)
> {
> /* Do all things that we don't want to know about outside this function
>*/
> }
>
> /* turn off all swap devices */
> int swapoff_all()
> {
> int index;
> struct swdevt *sp;
>
> for (sp = swdevt, index = 0; index < nswdev; index++, sp++)
> swapoff_one(sp);
> }
>
> So the swapoff() system call would call swapoff_one() and my code in
> kern/kern_swsuspend.c would call swapoff_all().
See swapoff(8), in particular the -a flag.
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
Re: nswap
On Mon, Jan 20, 2003 at 11:09:36AM -0800, David Schultz wrote:
> What exactly do you need to change about the swapoff interface?
> Unless you're trying to write a module, anything that's going to
> be invasive into the swap subsystem's data structures probably
> belongs in vm_swap.c.
Yes, that is also what I meant. We now have a swapoff() system call that
does all the work itself.
My idea was to split that up:
/* turn of swap device */
static int swapoff_one(struct swdevt *sp)
{
/* Do all things that we don't want to know about outside this function
*/
}
/* turn off all swap devices */
int swapoff_all()
{
int index;
struct swdevt *sp;
for (sp = swdevt, index = 0; index < nswdev; index++, sp++)
swapoff_one(sp);
}
So the swapoff() system call would call swapoff_one() and my code in
kern/kern_swsuspend.c would call swapoff_all().
Mark
--
Mark SantcroosRIPE Network Coordination Centre
http://www.ripe.net/home/mark/New Projects Group/TTM
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
Re: nswap
Thus spake Mark Santcroos <[EMAIL PROTECTED]>: > On Fri, Jan 17, 2003 at 03:27:29PM -0800, David Schultz wrote: > > Thus spake Mark Santcroos <[EMAIL PROTECTED]>: > > > in : > > > extern int nswap; /* size of swap space */ > > > > > > in : > > > static int nswap; /* first block after the interleaved devs */ > > > > > > Is the extern pointing to this variable? (It seems so, don't see any other > > > such variable in the three) > > > If so, is there any problem with making nswap non-static? > > > > It's a constant that is only relevant to the management of the > > swap allocation bitmap, so it is properly static. It shouldn't be > > declared in sys/systm.h. > > I need to get access to the swap interface. More precisely, I want to turn > it off. > > Either some of the values and functions need to be made global or the > interface should be changed. > > I need this for software suspending on which I am working. (Page in processes > and turn off swap devices before I write out the memory to it) > > Is it worth my time to produce patches that change the swapoff interface? What exactly do you need to change about the swapoff interface? Unless you're trying to write a module, anything that's going to be invasive into the swap subsystem's data structures probably belongs in vm_swap.c. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: nswap
On Fri, Jan 17, 2003 at 03:27:29PM -0800, David Schultz wrote: > Thus spake Mark Santcroos <[EMAIL PROTECTED]>: > > in : > > extern int nswap; /* size of swap space */ > > > > in : > > static int nswap; /* first block after the interleaved devs */ > > > > Is the extern pointing to this variable? (It seems so, don't see any other > > such variable in the three) > > If so, is there any problem with making nswap non-static? > > It's a constant that is only relevant to the management of the > swap allocation bitmap, so it is properly static. It shouldn't be > declared in sys/systm.h. I need to get access to the swap interface. More precisely, I want to turn it off. Either some of the values and functions need to be made global or the interface should be changed. I need this for software suspending on which I am working. (Page in processes and turn off swap devices before I write out the memory to it) Is it worth my time to produce patches that change the swapoff interface? Mark -- Mark SantcroosRIPE Network Coordination Centre http://www.ripe.net/home/mark/New Projects Group/TTM To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: nswap
Thus spake Mark Santcroos <[EMAIL PROTECTED]>: > in : > extern int nswap; /* size of swap space */ > > in : > static int nswap; /* first block after the interleaved devs */ > > Is the extern pointing to this variable? (It seems so, don't see any other > such variable in the three) > If so, is there any problem with making nswap non-static? It's a constant that is only relevant to the management of the swap allocation bitmap, so it is properly static. It shouldn't be declared in sys/systm.h. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
nswap
in : extern int nswap; /* size of swap space */ in : static int nswap; /* first block after the interleaved devs */ Is the extern pointing to this variable? (It seems so, don't see any other such variable in the three) If so, is there any problem with making nswap non-static? Thanks Mark -- Mark SantcroosRIPE Network Coordination Centre http://www.ripe.net/home/mark/New Projects Group/TTM To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
