Re: [PATCH 2 of 3] [KVM] Add DCR access information to struct kvm_run
On Mon, Apr 07, 2008 at 03:53:33PM -0500, Hollis Blanchard wrote: > 1 file changed, 7 insertions(+) > include/linux/kvm.h |7 +++ > > > Device Control Registers are essentially another address space found on > PowerPC > 4xx processors, analogous to PIO on x86. DCRs are always 32 bits, and are > identified by a 32-bit number. Well... 10-bit, actually. -- David Gibson| I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
Re: [PATCH 2 of 3] [KVM] Add DCR access information to struct kvm_run
On Monday 07 April 2008 20:11:28 David Gibson wrote: > On Mon, Apr 07, 2008 at 03:53:33PM -0500, Hollis Blanchard wrote: > > 1 file changed, 7 insertions(+) > > include/linux/kvm.h |7 +++ > > > > > > Device Control Registers are essentially another address space found on > > PowerPC 4xx processors, analogous to PIO on x86. DCRs are always 32 bits, > > and are identified by a 32-bit number. > > Well... 10-bit, actually. The mtdcrux description in the ppc440x6 user manual says the following: Let the contents of register RA denote a Device Control Register. The contents of GPR[RS] are placed into the designated Device Control Register. I take that to mean that we must worry about 32 bits worth of DCR numbers. Perhaps I should say "no more than" rather than "always". -- Hollis Blanchard IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
Re: [PATCH 2 of 3] [KVM] Add DCR access information to struct kvm_run
On Mon, Apr 07, 2008 at 10:25:32PM -0500, Hollis Blanchard wrote: > On Monday 07 April 2008 20:11:28 David Gibson wrote: > > On Mon, Apr 07, 2008 at 03:53:33PM -0500, Hollis Blanchard wrote: > > > 1 file changed, 7 insertions(+) > > > include/linux/kvm.h |7 +++ > > > > > > > > > Device Control Registers are essentially another address space found on > > > PowerPC 4xx processors, analogous to PIO on x86. DCRs are always 32 bits, > > > and are identified by a 32-bit number. > > > > Well... 10-bit, actually. > > The mtdcrux description in the ppc440x6 user manual says the following: > > Let the contents of register RA denote a Device Control Register. > The contents of GPR[RS] are placed into the designated Device Control > Register. > > I take that to mean that we must worry about 32 bits worth of DCR numbers. > Perhaps I should say "no more than" rather than "always". I think that's less misleading. mtdcrux is very new, anything which only has the mtdcr instruction certainly can't take DCR numbers above 10 bits, and I would expect that even on chips with mtdcrux the DCR bus is probably still only 10-bits, although it could be extended. -- David Gibson| I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
Re: [PATCH 2 of 3] [KVM] Add DCR access information to struct kvm_run
On Monday 07 April 2008 22:54:41 David Gibson wrote: > On Mon, Apr 07, 2008 at 10:25:32PM -0500, Hollis Blanchard wrote: > > On Monday 07 April 2008 20:11:28 David Gibson wrote: > > > On Mon, Apr 07, 2008 at 03:53:33PM -0500, Hollis Blanchard wrote: > > > > 1 file changed, 7 insertions(+) > > > > include/linux/kvm.h |7 +++ > > > > > > > > > > > > Device Control Registers are essentially another address space found > > > > on PowerPC 4xx processors, analogous to PIO on x86. DCRs are always > > > > 32 bits, and are identified by a 32-bit number. > > > > > > Well... 10-bit, actually. > > > > The mtdcrux description in the ppc440x6 user manual says the following: > > > > Let the contents of register RA denote a Device Control Register. > > The contents of GPR[RS] are placed into the designated Device Control > > Register. > > > > I take that to mean that we must worry about 32 bits worth of DCR > > numbers. Perhaps I should say "no more than" rather than "always". > > I think that's less misleading. mtdcrux is very new, anything which > only has the mtdcr instruction certainly can't take DCR numbers above > 10 bits, and I would expect that even on chips with mtdcrux the DCR > bus is probably still only 10-bits, although it could be extended. We're defining a kernel/userspace interface here, and since the hardware is capable of 32-bit DCR numbers, I don't think it makes any sense to not support that. Also, we would just end up placing that number into a u32 anyways, so... :) -- Hollis Blanchard IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
Re: [PATCH 2 of 3] [KVM] Add DCR access information to struct kvm_run
On Mon, Apr 07, 2008 at 11:06:10PM -0500, Hollis Blanchard wrote: > On Monday 07 April 2008 22:54:41 David Gibson wrote: > > On Mon, Apr 07, 2008 at 10:25:32PM -0500, Hollis Blanchard wrote: > > > On Monday 07 April 2008 20:11:28 David Gibson wrote: > > > > On Mon, Apr 07, 2008 at 03:53:33PM -0500, Hollis Blanchard wrote: > > > > > 1 file changed, 7 insertions(+) > > > > > include/linux/kvm.h |7 +++ > > > > > > > > > > > > > > > Device Control Registers are essentially another address space found > > > > > on PowerPC 4xx processors, analogous to PIO on x86. DCRs are always > > > > > 32 bits, and are identified by a 32-bit number. > > > > > > > > Well... 10-bit, actually. > > > > > > The mtdcrux description in the ppc440x6 user manual says the following: > > > > > > Let the contents of register RA denote a Device Control Register. > > > The contents of GPR[RS] are placed into the designated Device Control > > > Register. > > > > > > I take that to mean that we must worry about 32 bits worth of DCR > > > numbers. Perhaps I should say "no more than" rather than "always". > > > > I think that's less misleading. mtdcrux is very new, anything which > > only has the mtdcr instruction certainly can't take DCR numbers above > > 10 bits, and I would expect that even on chips with mtdcrux the DCR > > bus is probably still only 10-bits, although it could be extended. > > We're defining a kernel/userspace interface here, and since the hardware is > capable of 32-bit DCR numbers, I don't think it makes any sense to not > support that. Also, we would just end up placing that number into a u32 > anyways, so... :) Oh, of course you should represent it as a u32 and support 32-bit addresses, it's only the patch comment I'm objecting to. -- David Gibson| I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
Re: [PATCH 2 of 3] [KVM] Add DCR access information to struct kvm_run
On Tue, 8 Apr 2008 13:54:41 +1000 David Gibson <[EMAIL PROTECTED]> wrote: > On Mon, Apr 07, 2008 at 10:25:32PM -0500, Hollis Blanchard wrote: > > On Monday 07 April 2008 20:11:28 David Gibson wrote: > > > On Mon, Apr 07, 2008 at 03:53:33PM -0500, Hollis Blanchard wrote: > > > > 1 file changed, 7 insertions(+) > > > > include/linux/kvm.h |7 +++ > > > > > > > > > > > > Device Control Registers are essentially another address space found on > > > > PowerPC 4xx processors, analogous to PIO on x86. DCRs are always 32 > > > > bits, > > > > and are identified by a 32-bit number. > > > > > > Well... 10-bit, actually. > > > > The mtdcrux description in the ppc440x6 user manual says the following: > > > > Let the contents of register RA denote a Device Control Register. > > The contents of GPR[RS] are placed into the designated Device Control > > Register. > > > > I take that to mean that we must worry about 32 bits worth of DCR numbers. > > Perhaps I should say "no more than" rather than "always". > > I think that's less misleading. mtdcrux is very new, anything which > only has the mtdcr instruction certainly can't take DCR numbers above > 10 bits, and I would expect that even on chips with mtdcrux the DCR > bus is probably still only 10-bits, although it could be extended. http://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/C94B06BE313211B887257110006EFFBD/$file/460migrate.pdf page 4. "DCR Address Space Increased to 32 bits". I realize that the above is for 460 cores, but I would not be surprised at all if that shows up in a future 440 core. 440x6 already seems to be a conglomeration of some of the features 460 has. josh ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
