[BKPATCH] allow pci primary peer busses to have parents

[BKPATCH] allow pci primary peer busses to have parents

[James Bottomley]

Now that the generic device model allows a coherent bus tree to be built, 
there are certain architectures that hang the PCI primary busses off another 
bus (I need this patch for parisc, but I'm sure infiniband would have similar 
issues).

This patch allows the allocation of parented pci primary peer busses so that 
they slot correctly into the device model.

James

You can import this changeset into BK by piping this whole message to:
'| bk receive [path to repository]' or apply the patch as usual.

===================================================================


ChangeSet@1.927, 2002-12-03 09:28:37-06:00, jejb@raven.il.steeleye.com
  allow pci primary busses to have parents in the device model


 drivers/pci/probe.c |    7 ++++---
 include/linux/pci.h |   12 ++++++++++--
 2 files changed, 14 insertions(+), 5 deletions(-)


diff -Nru a/drivers/pci/probe.c b/drivers/pci/probe.c
--- a/drivers/pci/probe.c	Wed Dec  4 11:11:38 2002
+++ b/drivers/pci/probe.c	Wed Dec  4 11:11:38 2002
@@ -547,7 +547,7 @@
 	return 0;
 }
 
-struct pci_bus * __devinit pci_alloc_primary_bus(int bus)
+struct pci_bus * __devinit pci_alloc_primary_bus_parented(struct device 
*parent, int bus)
 {
 	struct pci_bus *b;
 
@@ -566,6 +566,7 @@
 	memset(b->dev,0,sizeof(*(b->dev)));
 	sprintf(b->dev->bus_id,"pci%d",bus);
 	strcpy(b->dev->name,"Host/PCI Bridge");
+	b->dev->parent = parent;
 	device_register(b->dev);
 
 	b->number = b->secondary = bus;
@@ -574,9 +575,9 @@
 	return b;
 }
 
-struct pci_bus * __devinit pci_scan_bus(int bus, struct pci_ops *ops, void 
*sysdata)
+struct pci_bus * __devinit pci_scan_bus_parented(struct device *parent, int 
bus, struct pci_ops *ops, void *sysdata)
 {
-	struct pci_bus *b = pci_alloc_primary_bus(bus);
+	struct pci_bus *b = pci_alloc_primary_bus_parented(parent, bus);
 	if (b) {
 		b->sysdata = sysdata;
 		b->ops = ops;
diff -Nru a/include/linux/pci.h b/include/linux/pci.h
--- a/include/linux/pci.h	Wed Dec  4 11:11:38 2002
+++ b/include/linux/pci.h	Wed Dec  4 11:11:38 2002
@@ -520,8 +520,16 @@
 /* Generic PCI functions used internally */
 
 int pci_bus_exists(const struct list_head *list, int nr);
-struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops, void *sysdata);
-struct pci_bus *pci_alloc_primary_bus(int bus);
+struct pci_bus *pci_scan_bus_parented(struct device *parent, int bus, struct 
pci_ops *ops, void *sysdata);
+static inline struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops, void 
*sysdata)
+{
+	return pci_scan_bus_parented(NULL, bus, ops, sysdata);
+}
+struct pci_bus *pci_alloc_primary_bus_parented(struct device * parent, int 
bus);
+static inline struct pci_bus *pci_alloc_primary_bus(int bus)
+{
+	return pci_alloc_primary_bus_parented(NULL, bus);
+}
 struct pci_dev *pci_scan_slot(struct pci_dev *temp);
 int pci_proc_attach_device(struct pci_dev *dev);
 int pci_proc_detach_device(struct pci_dev *dev);

===================================================================


This BitKeeper patch contains the following changesets:
1.927
## Wrapped with gzip_uu ##


begin 664 bkpatch568
M'XL(`$HW[CT``\U6:V_3,!3]7/^**_%EC"7U(\].G<9+@)A@&MHGA"HW,6NV
M-*ELIS`1_CO7R>C&:,LV@413Q9)S?>\YQ_<X>02G1NG1X%R=3\DC>%T;.QIH
MN5257Y2^L4J5ZE+Y63W'IR=UC4^'LWJNAF[!\-G;X2(KO(74JK)%=>9Q/_1#
M2C#V6-IL!DNES6C`?+&:L9<+-1J<O'QU>O3TA)#Q&)[/9'6F/B@+XS&QM5[*
M,C>'TL[*NO*MEI69*RL=A'85VG)*.5XABP4-HY9%-(C;C.6,R8"IG/(@B8+K
M;`[RUER,\9AS(<*DI0D5G+P`YJ<\!LJ'C`^I`)J.>#(2L4>C$:7@^!^N%0J>
M</`H>09_E\MSDH$LR_H+H.2PT,5<ZDN8-L8H@Z5@AEB@WPD#105VIB!7RR)3
M,*]S59*W$"%-<GPM./'N^2.$2DH.,&$V4^6A46>?95-:OS9YZ=?ZK,UUX;;<
M=<5PH>LI"M+3"Q@6YZ%(VU"@OFV:Y"J@4\DRFH8AS;;HN3$IXU2PD,=AW*)"
M,4-@VR4OJJQL<C4LBZKYZM+YLYOBIP%M&1-QU$9<)(E4/$]#F49!N@W=IJ0W
MT-&`1[SK]3547-?_?3U)45T<GC=:&E-D/K;%A3\WC:^;CS\E^K195X3.XH`A
M<I'$-.Z\P.FO5HA'X5VL$(`G_D<K]`WS'CS]I?MC;Q^OVYT'6.0%Z@^,O.D'
M8W6360=T@@!A%R83!Z4J^CE'(YM<47`1DQZWRG>N5EX!W^WG]Y"/=50?8X$H
MP0*#J7>`,=Y!'P#C*^;[""2..R#=\`<@)I/5?>KOP8V$]0(3XFT/EG61PZZY
M-+FT\K&#D/80W#"XC6'JT&Y7X6=91WF_\]`:P_WYS?%@Z]_%2!N/`,8Y2VD0
M1*U(61)U1@K%@XS$*'C\OW22.]MN&6F-(`\R$A?`L7EX`(S^UL#_K&OWL92T
M18:+D("";85W[F6';V2@E6UTM<%R[TZ/CO;Z=-WR:T3?U]*_^^D!MX^/NY#\
H+?W.ZO#YE<D6'"M*'8G59R"^[[(+T\S'TSQA:(XI^0$X3;EMB@H`````
`
end

Re: [BKPATCH] allow pci primary peer busses to have parents

[Ivan Kokshaysky]

On Wed, Dec 04, 2002 at 11:18:24AM -0600, James Bottomley wrote:
> Now that the generic device model allows a coherent bus tree to be built

Unfortunately it doesn't. Currently those legacy, PnP, EISA devices all have
virtual parents, which has nothing to do with reality. Modern systems
(including most PCs) hang these buses off PCI bus using PCI-to-{E}ISA
bridge. Such systems must be able to register these buses upon
discovery of the ISA bridges (from pci layer), and use them as a
parent device for legacy/isa/pnp stuff. This will be absolutely required
if DMA operations are moved from pci_dev to the generic device.

> -struct pci_bus * __devinit pci_scan_bus(int bus, struct pci_ops *ops, void 
> *sysdata)
> +struct pci_bus * __devinit pci_scan_bus_parented(struct device *parent, int 
> bus, struct pci_ops *ops, void *sysdata)
>  {
> -	struct pci_bus *b = pci_alloc_primary_bus(bus);
> +	struct pci_bus *b = pci_alloc_primary_bus_parented(parent, bus);
>  	if (b) {
>  		b->sysdata = sysdata;
>  		b->ops = ops;

The `sysdata' arg already contains info about parent host-to-pci controller
on many platforms. I don't think that we need to duplicate it with
another one.
I was thinking about something like this instead of `sysdata':

struct io_controller {		/* Level 0 I/O controller */
	... arch specific fields ...;
	... generic fields ...;	/* like `index' */
	struct device dev;
}

Ivan.

Re: [BKPATCH] allow pci primary peer busses to have parents

[James Bottomley]

On Wed, Dec 04, 2002 at 11:18:24AM -0600, James Bottomley wrote:
> Now that the generic device model allows a coherent bus tree to be built

ink@jurassic.park.msu.ru said:
> Unfortunately it doesn't.

OK, consider the phrase for architectures with subordinate PCI busses added.

ink@jurassic.park.msu.ru said:
>  Currently those legacy, PnP, EISA devices all have virtual parents,
> which has nothing to do with reality. Modern systems (including most
> PCs) hang these buses off PCI bus using PCI-to-{E}ISA bridge. Such
> systems must be able to register these buses upon discovery of the ISA
> bridges (from pci layer), and use them as a parent device for legacy/
> isa/pnp stuff. This will be absolutely required if DMA operations are
> moved from pci_dev to the generic device. 

Well, we are moving in this direction.  I've already done the conversion for 
MCA.  Marc Zyngier has done it for EISA.  I believe someone is looking at PnP 
ISA.  ISA, as a non-probe'able bus fits into the legacy bus scheme anyway.

> The `sysdata' arg already contains info about parent host-to-pci
> controller on many platforms. I don't think that we need to duplicate
> it with another one. I was thinking about something like this instead
> of `sysdata': 

That's PCI specific.  We need a coherent tree in the generic model.  To do 
this, the PCI parent information has to be available just using the struct 
device, without having to cast it to pci_dev and look at pci specific fields.

This is a simplification requirement for machines whose IOMMUs lie on other 
bus types above the PCI busses.  You have to be able to walk up the device 
tree until you find the IOMMU.  Since you're sharing the implementation with 
the non-PCI busses, you need to be able to do this in a generic manner.

James

Re: [BKPATCH] allow pci primary peer busses to have parents

[Ivan Kokshaysky]

On Thu, Dec 05, 2002 at 09:33:29AM -0600, James Bottomley wrote:
> ink@jurassic.park.msu.ru said:
> > isa/pnp stuff. This will be absolutely required if DMA operations are
> > moved from pci_dev to the generic device. 
> 
> Well, we are moving in this direction.  I've already done the conversion for 
> MCA.  Marc Zyngier has done it for EISA.  I believe someone is looking at PnP 
> ISA.  ISA, as a non-probe'able bus fits into the legacy bus scheme anyway.

Nice to know. :-)
Current approach with initcalls doesn't work - basically, only architecture
specific code knows what bus types are primary and therefore should be
initialized first. IOW, it would be good to have something like you've
suggested for PCI for any bus type:

XXX_bus_init(struct device *parent, int busnum)

> > The `sysdata' arg already contains info about parent host-to-pci
> > controller on many platforms. I don't think that we need to duplicate
> > it with another one. I was thinking about something like this instead
> > of `sysdata': 
> 
> That's PCI specific.

Actually it isn't. It just happens to be that sysdata == pci_controller
on pci-based machines. However, these structures have very little to
do with PCI - it's all about IOMMUs, various address ranges and other
host-specific data.

>  We need a coherent tree in the generic model.  To do 
> this, the PCI parent information has to be available just using the struct 
> device, without having to cast it to pci_dev and look at pci specific fields.

Of course.

> This is a simplification requirement for machines whose IOMMUs lie on other 
> bus types above the PCI busses.

That's why I suggested `io_controller' name, but I can live with just
`sysdata' as well. :-)

> You have to be able to walk up the device 
> tree until you find the IOMMU.  Since you're sharing the implementation with 
> the non-PCI busses, you need to be able to do this in a generic manner.

Right, but walking up the entire tree every time is rather painful.
Things like pci^H^H^Hdma_map_{single,sg} are supposed to be fast, so I'd
like to gather IOMMU and other info directly from struct device * passed
as argument to these functions.

Ivan.