Bug fixes for 2.6.20
Subsystems (2)
If anyone has a patch that they need to have in 2.6.20 in order to fix a bug or make some existing code compile, please let me know. I already have a patch from Michal Ostrowski, two patches from Ben H. and a set from Stephen Rothwell in my queue. I don't know what's needed for any of the embedded boards, though.
Paul.
PATCH 1/1 Probe Efika platform before CHRP.
The Efika matches chrp_probe() too, so put its own probe first to make sure we get it right in a multiplatform build.
arch/powerpc/platforms/Makefile renamed from arch/powerpc/platforms/Makefile~ +1 -1
@@ -5,9 +5,9 @@ ifeq ($(CONFIG_PPC64),y) obj-$(CONFIG_PPC_PMAC) += powermac/ endif endif +obj-$(CONFIG_PPC_MPC52xx) += 52xx/ obj-$(CONFIG_PPC_CHRP) += chrp/ obj-$(CONFIG_4xx) += 4xx/ -obj-$(CONFIG_PPC_MPC52xx) += 52xx/ obj-$(CONFIG_PPC_83xx) += 83xx/ obj-$(CONFIG_PPC_85xx) += 85xx/ obj-$(CONFIG_PPC_86xx) += 86xx/
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 2 lines · David Woodhouse
The Efika matches
chrp_probe()too, so put its own probe first to make sure we get it right in a multiplatform build.
Hrm... I told them to remove that from their device-tree ... dammit !
Quoted 16 lines
Signed-off-by David Woodhouse <[email protected]>arch/powerpc/platforms/Makefile renamed from arch/powerpc/platforms/Makefile~ +1
-1@@ -5,9 +5,9 @@ ifeq ($(CONFIG_PPC64),y) obj-$(CONFIG_PPC_PMAC) += powermac/ endif endif +obj-$(CONFIG_PPC_MPC52xx) += 52xx/ obj-$(CONFIG_PPC_CHRP) += chrp/ obj-$(CONFIG_4xx) += 4xx/ -obj-$(CONFIG_PPC_MPC52xx) += 52xx/ obj-$(CONFIG_PPC_83xx) += 83xx/ obj-$(CONFIG_PPC_85xx) += 85xx/ obj-$(CONFIG_PPC_86xx) += 86xx/
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 8 lines · Benjamin Herrenschmidt
Quoted 3 lines · David Woodhouse
The Efika matches
chrp_probe()too, so put its own probe first to make sure we get it right in a multiplatform build.Hrm... I told them to remove that from their device-tree ... dammit !
Yes, I told them to do some other changes to their device-tree, but no luck so far.
I think I'll end up with a bunch of device-tree fix in the efika probe function. (Either
that, or force the user to have a nvramrc for the firmware that makes the appropriate
changes, but that's kindof an hassle for the user)
That won't solve the problem with chrp though.
I've seen that pseries will also match chrp. And in the pseries probe function,
the Cell blades are explicitely 'refused'.
Maybe instead of putting 52xx first, it would be "better" to put chrp and pseries
last as they seem more "generic". So that anything that match before that has
priority.
Sylvain
Quoted 24 lines
Quoted 7 lines
Signed-off-by David Woodhouse <[email protected]>arch/powerpc/platforms/Makefile ifeq ($(CONFIG_PPC64),y) · line 13
Quoted 10 lines
arch/powerpc/platforms/Makefile renamed from arch/powerpc/platforms/Makefile~ +1
-1@@ -5,9 +5,9 @@ ifeq ($(CONFIG_PPC64),y) obj-$(CONFIG_PPC_PMAC) += powermac/ endif endif +obj-$(CONFIG_PPC_MPC52xx) += 52xx/ obj-$(CONFIG_PPC_CHRP) += chrp/ obj-$(CONFIG_4xx) += 4xx/ -obj-$(CONFIG_PPC_MPC52xx) += 52xx/ obj-$(CONFIG_PPC_83xx) += 83xx/ obj-$(CONFIG_PPC_85xx) += 85xx/ obj-$(CONFIG_PPC_86xx) += 86xx/_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-dev
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 11 lines · Sylvain Munaut
Quoted 7 lines · Benjamin Herrenschmidt
Quoted 3 lines · David Woodhouse
The Efika matches
chrp_probe()too, so put its own probe first to make sure we get it right in a multiplatform build.Hrm... I told them to remove that from their device-tree ... dammit !
Yes, I told them to do some other changes to their device-tree, but no luck so far.
OK, let's drop the patch and assume the device-tree will be fixed shortly. In the meantime, users should be able to work around it by hacking the device-tree from the OF prompt -- which we currently have to do for other things in the Efika device-tree anyway.
I'm happy enough for Fedora to require the fixed firmware, certainly.
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 11 lines · David Woodhouse
Quoted 5 lines · Sylvain Munaut
Quoted 1 line · Benjamin Herrenschmidt
Hrm... I told them to remove that from their device-tree ... dammit !
Yes, I told them to do some other changes to their device-tree, but no luck so far.
OK, let's drop the patch and assume the device-tree will be fixed shortly. In the meantime, users should be able to work around it by hacking the device-tree from the OF prompt -- which we currently have to do for other things in the Efika device-tree anyway.
This will do for the time being; but I'm not expecting any changes to be made to the firmware (however, Sven is making the attempt). In the long term, I think we are going to have to live with either OF fixups; or preventing CHRP from matching the efika and adding other fixups in the Efika's probe function. I like Sylvain's suggestion of moving CHRP and pseries down in the link order; but then again, I'm neither a CHRP or pseries user. :)
Cheers,
g.
Signature · 4 lines
Grant Likely, B.Sc. P.Eng. Secret Lab Technologies Ltd. [email protected] (403) 399-0195
Re: [PATCH] Probe Efika platform before CHRP.
In reply to David WoodhouseQuoted 3 lines
OK, let's drop the patch and assume the device-tree will be fixed shortly.
No, keep the patch for now. I'm afraid they aren't willing to fix their device-tree.
Quoted 6 lines
In the meantime, users should be able to work around it by hacking the device-tree from the OF prompt -- which we currently have to do for other things in the Efika device-tree anyway.
I'm happy enough for Fedora to require the fixed firmware, certainly.
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 7 lines · Benjamin Herrenschmidt
Quoted 3 lines
OK, let's drop the patch and assume the device-tree will be fixed shortly.
No, keep the patch for now. I'm afraid they aren't willing to fix their device-tree.
Ok,
but please, put a comment near it would be nice if we remember why we put it there.
Sylvain
Quoted 10 lines
Quoted 8 lines
In the meantime, users should be able to work around it by hacking the device-tree from the OF prompt -- which we currently have to do for other things in the Efika device-tree anyway.
I'm happy enough for Fedora to require the fixed firmware, certainly.
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Benjamin HerrenschmidtQuoted 5 lines · Benjamin Herrenschmidt
Quoted 2 lines
OK, let's drop the patch and assume the device-tree will be fixed shortly.
No, keep the patch for now. I'm afraid they aren't willing to fix their device-tree.
They need to fix their firmware anyway, because it's screwed in other ways too -- they have an off-by-one error in their handling of partition numbers, so when yaboot tries to read a partition table it actually gets the first sector of the first partition, and it all goes downhill from there. That one was even _fixed_ in the Pegasos firmware last year, but it's come back in Efika -- and we _can't_ work around it in the kernel.
For Fedora, we're just going to wait for the fixed firmware; we can't sensibly support it as-is.
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 14 lines · David Woodhouse
Quoted 5 lines · Benjamin Herrenschmidt
Quoted 2 lines
OK, let's drop the patch and assume the device-tree will be fixed shortly.
No, keep the patch for now. I'm afraid they aren't willing to fix their device-tree.
They need to fix their firmware anyway, because it's screwed in other ways too -- they have an off-by-one error in their handling of partition numbers, so when yaboot tries to read a partition table it actually gets the first sector of the first partition, and it all goes downhill from there. That one was even _fixed_ in the Pegasos firmware last year, but it's come back in Efika -- and we _can't_ work around it in the kernel.
Quoted 2 lines
For Fedora, we're just going to wait for the fixed firmware; we can't sensibly support it as-is.
Ok. Well, Nicolas Det have unfortunately repeately said "we won't fix the firmware" so it's bloody annoying but if you can apply that sort of pressure on from a distro, that's good.
I've been very annoyed from the beginning with the refusal to discuss the content of the device-tree publically before it was too late, the firmware "in production" and "not fixable" (Nicolas words).
Ben.
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 2 lines · Benjamin Herrenschmidt
Ok. Well, Nicolas Det have unfortunately repeately said "we won't fix the firmware"
Really? Why on earth not? We're talking about simple and obvious fixes, and we have no reason to keep it broken for "compatibility", surely?
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 6 lines · David Woodhouse
Quoted 2 lines · Benjamin Herrenschmidt
Ok. Well, Nicolas Det have unfortunately repeately said "we won't fix the firmware"
Really? Why on earth not? We're talking about simple and obvious fixes, and we have no reason to keep it broken for "compatibility", surely?
Really don't know, especially considering that the firmware did get changed during the whole process to modify the irq mapping format.
Cheers,
g.
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 3 lines · Matt Sealey
One basic problem is we're talking about changing Efika to adapt to the definitions, whims and determinations of the *Linux* community.
As opposed to the Linux community being asked to adapt to the definitions, whims and determinations of a single corporate entity? :-)
Quoted 2 lines
What if NetBSD, FreeBSD, QNX and WindRiver disagree with you guys?
I'm pretty sure that Linux is the only OS which is requiring a device tree for booting for new ports, and I'm also pretty sure that the Efika is the only 52xx board with open firmware on it. If I'm right about this; then for the vast majority of 52xx boards, a device tree is only needed if it is booting Linux, and therefore it is up to the Linux community to define what it needs to look like. It also makes the Efika the exception, not the rule (as it's the only board which will pass the device tree to non-Linux OSes)
Quoted 4 lines
Who do we change for?
You might understand, we all think you're being just a TAD unreasonable on this matter.
I'm not trying to say "Change Dammit!". I'm trying to say, "This is what I think it should look like; where should it go from here". So far I feel like I've been told: "we've done it this way, deal with it"
Quoted 4 lines
Surely Linux should detect all available and present nodes, compatibility flags and so on, and not simply tell vendors to "shape up your firmware because we refuse to make Linux support anything we did not define and impose on you"?
Early on when we were trying to draft a standard for 52xx device trees (mpc52xx-device-tree-bindings.txt), bplan refused to disclose what the Efika device tree layout looked like. In the absence of any input from bplan, a draft 52xx device tree spec was written. Mostly, the draft spec followed the conventions already established for other powerpc SoC devices (see booting-without-of.txt; and Yes, I realize that the Efika has real-of; but I'm just referring to the naming convention portions of that document).
At a later date, Nicolas made the statement that the firmware was set and there would be not changes to the device tree, so I better just copy what was done on the Efika.
I do not object to changing the definitions or what we are requesting for the device tree layout. I do object to being told, "we did it this way, so just copy it" after refusing to discuss any kind of device tree layout standard for the 52xx parts. As someone who is not getting paid to do this work, it does not particularly motivate me to want to help you out.
Quoted 4 lines
Specifically for the "sound" and "ac97" discrepency, the "sound" device type has some special implications for Open Firmware firmwares. It's set to ac97 as it is NOT an OF or CHRP standard "sound" node and should NOT be found as such in my personal opinion.
For the record; this has already been addressed:
http://patchwork.ozlabs.org/linuxppc/patch?id=8842
'ac97' was wrong, it should have been 'sound'
I do *NOT* want to claim that the proposed mpc52xx device tree binding is correct or final. My motivation is to see a consistent, well thought out device tree specification for all 52xx boards before too many vendors produce device trees that are utterly incompatible with one another (due to inconsistent naming).
Quoted 5 lines
Would you mind giving me a list of exactly WHAT you think is wrong (the patch doesn't define what is broken, it only blindly throws in the fixes with no explanation of why in comments or documentation..) and why we should fix it, and I can go through it with you, since I understand Nico is not up to discussing the matter being very busy and all :D
There are two issues here:
1. consistent mpc52xx soc device naming so that the appropriate compatible device driver can be selected; two different 52xx devices exist at the moment, and the Efika seems to report them as the wrong one (ie. 5200 instead of 5200b). Plus there are rumblings from Freescale to be bringing out more 52xx devices. These are the issues that mpc52xx-device-tree-bindings.txt was written to address; and that document also contains the rational for the naming convention. As it stands right now, Sylvain's patch provides a workaround for these concerns, so the differences between the draft spec and the Efika are contained within a single source file.
Once again, I am not opposed to changing the draft; the goal of the document is to encourage consistency in board designs.
2. Dwmw2 says that reading the HD has an off-by-one showstopper bug which prevents Linux from reading the partition table (and that it is the same bug that was on the pegasos, but is now fixed).
I have not confirmed this issues as I don't have a working HD on my Efika. However, dwmw2 has stated that he cannot support the Efika in Fedora until this issue is resolved.
Cheers,
g.
Signature · 4 lines
Grant Likely, B.Sc. P.Eng. Secret Lab Technologies Ltd. [email protected] (403) 399-0195
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 3 lines · Grant Likely
2. Dwmw2 says that reading the HD has an off-by-one showstopper bug which prevents Linux from reading the partition table (and that it is the same bug that was on the pegasos, but is now fixed).
It's an off-by-one in partition table numbering. When we ask for "hd:0" (the whole disk) we get "hd:1", and so on. It screws yaboot's partition table scan, because when it asks for the first sector of the disk it's given the first sector of the first partition.
Something very similar to this already got fixed on Pegasos after I reported it -- and I didn't think it was _particularly_ far-fetched for me to expect the Efika firmware to be adapted to the "definitions, whims and determinations" of the CHRP specification too.
Quoted 3 lines
I have not confirmed this issues as I don't have a working HD on my Efika. However, dwmw2 has stated that he cannot support the Efika in Fedora until this issue is resolved.
"Cannot" is a strong word when you're talking about software -- and is almost always untrue. I think what I said was "can't sensibly", which makes a lot more sense.
I've got a locally-hacked yaboot which works around it enough to boot my system, but I'm not going to get far if I suggest that we ship a different copy of the boot loader just for Efika "because it's speshul". Especially when the bug was known so long ago.
I think the workaround on Pegasos was to use Amiga partitions, and to hack (perhaps even unintentionally) the Amiga partition code in yaboot to compensate for the brokenness. So the Amiga partition code probably won't work on CHRP-compatible machines, while the rest of the partition code won't work on Efika/oldPegasos. I baulked at the idea of continuing this hack by using Amiga partitions on Efika too, and preferred to put the thing back on the shelf and wait for it to be fixed -- especially as I'm away for the rest of the month anyway.
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 6 lines · David Woodhouse
Quoted 1 line · Grant Likely
... he cannot support the Efika in Fedora until this issue is resolved.
"Cannot" is a strong word when you're talking about software -- and is almost always untrue. I think what I said was "can't sensibly", which makes a lot more sense.
Heh, sorry. I should know better. :)
g.
Signature · 4 lines
Grant Likely, B.Sc. P.Eng. Secret Lab Technologies Ltd. [email protected] (403) 399-0195
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Grant LikelyQuoted 2 lines
Quoted 2 lines
What if NetBSD, FreeBSD, QNX and WindRiver disagree with you guys?
Note that the "compatible" modification we ask could simply be to add the one
we require at the start of your compatible list, so linux will work and the os using
the efika original will work as well.
Quoted 4 lines
Quoted 4 lines
Specifically for the "sound" and "ac97" discrepency, the "sound" device type has some special implications for Open Firmware firmwares. It's set to ac97 as it is NOT an OF or CHRP standard "sound" node and should NOT be found as such in my personal opinion.
This "sound" <> "ac97" thing is not fixed in the patch I posted recently nor in the nvramrc I sent to gentoo before that. The only visible thing was a a small comment buried in a very experimental driver.
As soon as nicolas told me "sound" was the standard, I said OK so be it.
But for example the "memory" type you give to the SRAM node is imho wrong. Because memory seems to be the standard type for "real" memory, the one that's gonna be used for the system ...
Quoted 1 line
There are two issues here:
I'd say you can even split a little more :
* Compatible / Type of nodes :
We need a naming schema that's coherent across nodes and will allow to support coherently different boards. Your naming scheme just doesn't provide that. The one proposed by Grant does. Now it's possible you can find even better and we're open to suggestion.
* Missing bits :
- The interrupts property of the ac97 node are just not there. This
interrupt
exist, it's hw and you can't just decide the driver can do without
it ...
* Bugs : I see three,
- The one dwmv2 mentionned
- I noticed the compatible properties had double \0 at the end instead of single ones.
- Incorrect init of the PSC2 for AC97 and in the tree since the "gpio" node is not present, the address for port config is not found in the device tree. The current work around is a ugly hardcoded stuff in the driver itself that will never be merged upstream.
Regards,
Sylvain
PS: In the end, a long nvramrc will do the trick ... so I don't care much for me,
I'll always know what I need to put in it.
Re: [PATCH] Probe Efika platform before CHRP.
Hi Syvain, Ben, and the others here on this list...
In the next 45-60 days we will offer new firmware for the EFIKA (and the Pegasos/ODW). We will define a new hardware specification outside of CHRP. In the meanwhile and afterwards, feel free to proceed as you determine best. We appreciate the support we have received in the past and the effort made more recently on the EFIKA.
The long term objective for Genesi is to provide a low cost, low power, highly reliable computing platform that can be flexibly applied to many uses. We hope to be selling the EFIKA at $99 by April (and sooner if possible). The 5200B is to be followed by another SoC with integrated graphic support at the same price (for the SoC). The new firmware will be inexorably linked to this new SoC in form and function. The EFIKA is a development platform that targets the system we will offer with the new SoC. We expect to be shipping the next version of the EFIKA before the end of 2007. The first diagram shown in the link that follows (the second image is what we hope for after that):
http://bbrv.blogspot.com/2006/10/wanting-what-you-have.html
Thanks again. Happy New Year. :)
Best regards,
R&B
Quoted 62 lines · Sylvain Munaut
Quoted 2 lines
Quoted 2 lines
What if NetBSD, FreeBSD, QNX and WindRiver disagree with you guys?
Note that the "compatible" modification we ask could simply be to add the one
we require at the start of your compatible list, so linux will work and the os using
the efika original will work as well.Quoted 4 lines
Quoted 4 lines
Specifically for the "sound" and "ac97" discrepency, the "sound" device type has some special implications for Open Firmware firmwares. It's set to ac97 as it is NOT an OF or CHRP standard "sound" node and should NOT be found as such in my personal opinion.
This "sound" <> "ac97" thing is not fixed in the patch I posted recently nor in the nvramrc I sent to gentoo before that. The only visible thing was a a small comment buried in a very experimental driver.
As soon as nicolas told me "sound" was the standard, I said OK so be it.But for example the "memory" type you give to the SRAM node is imho wrong. Because memory seems to be the standard type for "real" memory, the one that's gonna be used for the system ...
Quoted 1 line
There are two issues here:
I'd say you can even split a little more :
* Compatible / Type of nodes :
We need a naming schema that's coherent across nodes and will
allow to
support coherently different boards. Your naming scheme just
doesn't
provide that. The one proposed by Grant does. Now it's possible
you can
find even better and we're open to suggestion.* Missing bits :
- The interrupts property of the ac97 node are just not there. This
interrupt
exist, it's hw and you can't just decide the driver can do
without
it ...* Bugs : I see three,
- The one dwmv2 mentionned
- I noticed the compatible properties had double \0 at the end
instead
of single ones.
- Incorrect init of the PSC2 for AC97 and in the tree since the
"gpio"
node is not present, the address for port config is not found
in the
device tree. The current work around is a ugly hardcoded stuff in the driver itself that will never be merged upstream.Regards,
Sylvain
PS: In the end, a long nvramrc will do the trick ... so I don't care much for me,
I'll always know what I need to put in it.
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 2 lines · Raquel Velasco and Bill Buck
In the next 45-60 days we will offer new firmware for the EFIKA (and the Pegasos/ODW).
Was hoping to see this before the Fedora 7 release... ?
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
Absolutely everything will happen in April...
Quoted 9 lines · David Woodhouse
Quoted 2 lines · Raquel Velasco and Bill Buck
In the next 45-60 days we will offer new firmware for the EFIKA (and the Pegasos/ODW).
Was hoping to see this before the Fedora 7 release... ?
Signature · 1 line
dwmw2
Raquel and Bill
http://bbrv.blogspot.com
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 1 line · Raquel Velasco and Bill Buck
April...
OK, that should work for us. I've dusted off my Efika and made sure the Fedora kernel is in reasonable shape for it, but we really do need the partition handling sorted before it's going to be installable by normal people. Please let me know as soon as you've got a prerelease which I can try out.
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Raquel Velasco and Bill BuckQuoted 1 line · Raquel Velasco and Bill Buck
Absolutely everything will happen in April...
Which year? 2007 or 2010?
Quoted 7 lines · David Woodhouse
Quoted 6 lines · Raquel Velasco and Bill Buck
Quoted 2 lines
In the next 45-60 days we will offer new firmware for the EFIKA (and the Pegasos/ODW).
Was hoping to see this before the Fedora 7 release... ?
Re: [PATCH] Probe Efika platform before CHRP.
It is coming - Freescale is involved and this adds time to the release, but it is all for good reason.
Best regards,
R&B
Quoted 15 lines · Olaf Hering
Quoted 1 line · Raquel Velasco and Bill Buck
Absolutely everything will happen in April...
Which year? 2007 or 2010?
Quoted 7 lines · David Woodhouse
Quoted 6 lines · Raquel Velasco and Bill Buck
Quoted 2 lines
In the next 45-60 days we will offer new firmware for the EFIKA (and the Pegasos/ODW).
Was hoping to see this before the Fedora 7 release... ?
Raquel and Bill
http://bbrv.blogspot.com
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Grant LikelyQuoted 10 lines · Grant Likely
Quoted 3 lines · Matt Sealey
One basic problem is we're talking about changing Efika to adapt to the definitions, whims and determinations of the *Linux* community.
As opposed to the Linux community being asked to adapt to the definitions, whims and determinations of a single corporate entity? :-)
Quoted 1 line
Quoted 4 lines
Quoted 1 line
What if NetBSD, FreeBSD, QNX and WindRiver disagree with you guys?
I'm pretty sure that Linux is the only OS which is requiring a device tree for booting for new ports
NetBSD, FreeBSD will support the device tree for enumeration and the ports in progress already do, and they also keep the Client Interface open so they have a much better environment to work from than the Linux "cache" of the device tree. If the OF device is there and there is no driver in the OS, at least somehow (slowly) it could be used.
Quoted 1 line
Efika is the only 52xx board with open firmware on it.
Yes you're right.
At least from my point of view (I don't speak for Nico or Gerald here), here is my discussion on the subject.
#1: As the only Open Firmware platform on this chip, what *we* do matters here, we DO have to offer some stability in operating system support and that means once we fix a device tree it needs to stay fairly similar, which is exactly why we feel justified defining the tree standard for it :D
We do have to work with a lot more operating systems and vendors than just Linux, and a "booting without Open Firmware" standard which proves to solve the lack of device tree in U-Boot and other firmwares by making more work for firmware and Linux kernel porting harder, is not really something we are too concerned about. We hope we offer something a little better than "recompile your device tree file, embed it into the kernel, hope UBoot is the correct version, or recompile it, and flash and hope for the best". What we offer here is fairly unique; both in that we seem to be the only guys offering Open Firmware (especially a C-based one, OpenBIOS is 90% Forth which is a terrible lock-in), and on the MPC5200 (or PowerPC in general).
We also recognise OF and the device tree is flawed in some areas, RTAS isn't the best thing in the world (but it is totally misused in Linux, in places) so we are not BLIND to the need for change, we just think a few compatibility names and node device_type changes are not really where the real problems are.
#2: Personally I think naming device compatibility after the chip they are in is a ridiculous lock-in, also. You may end up with a device_type of ethernet, and compatibility flags for 100 SoC chip numbers. This is fairly stupid.
Isn't the e300 PVR and e300 SVR, or any other device identifier on the chip, a much better differentiator for drivers, than a named compatibility flag? These are present in the device tree (/cpu node?) no? I can't boot mine currently I blew up my power supply unfortunately.
I happen to know the next Freescale device won't be called 52xx so we're talking about adding another compatibility naming to it like 51xx or 55xx or 5xxx and changing everyone's firmware and Linux device trees again.
Or you could switch based on the processor! Cleaner, simpler, and doesn't need each device to have a specially unique name, only a vague "class" without some specific model number in it. If it's an SoC.. this is what the SVR is for!
#3: While the above is basically a case for an SoC node (device parent would have the SVR in it), this does not scale generically or flexibly to non-SoC platforms where devices are nevertheless highly integrated.
AMD's new processors with integrated CPU and memory controller, will not be classed as SoC's - how would you class them, really? Just a CPU with some devices builtin (aha!) I would say. With the Pegasos, we have the CPU, and the Marvell chip has Ethernet (not on PCI) and SRAM (also just mapped in as memory) but it is not an SoC.
We just don't want to clutter up the root node with devices, otherwise it would basically be more of a device shrubbery. The e300 SVR is also defined in the manual (as is the e500 equivalent, and e200 equivalent) as a CPU core feature, not a SoC-device feature, so this kind of differentiation and numbering belongs in the CPU node.
~
Those are MY problems with the device tree specification you have with "Booting Without Open Firmware". I am sure the other guys have other things to say.
As for real bugs in the firmware; we definitely have 3 or 4 I have noticed, be it the serial port not being changable baud rate from the firmware itself (" /builtin/serial:38400" io doesn't set the baud rate), AC97 is not configured (Sylvain hacked it though) even though we have a node that specifies it, disk numbering is DEFINITELY off (some facepalms going on at the office I think) even though we definitely already have a fix for it, among a lot of other niggly little things like the IrDA port (not set up, but I really think PSC6 should be configurable from a firmware variable because it is very multipurpose)
I would expect that it is so much easier, even if it does not strike you as too "clean", to forget about naming compatibility nodes here, and concentrate on the real showstoppers. I would consider the AC97 PSC not being set up, the IrDA PSC not being set up, the disk numbering and any problems we're having with USB devices (argh) to be real showstoppers.
We are working on a new firmware which will provide a lot of new functionality in the coming month or two. I am not sure anything that is merely based on principles or cleanliness of Linux code, to be changed. I know it's not nice to look at your own code and see a bunch of hacks for some hardware in there, but Linux and other OS are littered with these already, you cannot avoid a bug or two, or a discrepency in something else, and considering there are more important things in the design, and in life in general, maybe it is best just to use what you've got to work with (which is, unfortunately, what we gave you on the initial board..)?
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 6 lines
(especially a C-based one, OpenBIOS
is 90% Forth which is a terrible lock-in)Isn't the e300 PVR and e300 SVR, or any other device identifier on the chip, a much better differentiator for drivers, than a named compatibility flag?
Only 7 days into the new year, and already we get some great entries for best-joke-of-2007! Too bad they're factually incorrect.
Segher
Re: [PATCH] Probe Efika platform before CHRP.
Why is it so funny? I'd love to be enlightened rather than just plain insulted about it.
I don't think any of that can be taken into account by simply giving things *names* and then having a 3-week discussion and committee hearing about how someone gave it a DIFFERENT name. It is even more strange to me, that nobody HAS a standard name for the devices on the MPC5200B, yet we are arguing about what would look nicer in the Linux source code?
Small example:
mpc5200b-fec, mpc5200-fec, mpc52xx-fec doesn't tell you anything, and then with the 5500 or 5120 or something, you then have worked out you named your original devices wrong, and now need to make it compatible with those? Do you switch features on and off based on a string comparison? What if a certain chip revision has a bug you need to work around (there are plenty in the original MPC5200!), is that meant to be encoded in the "compatible" property too, somehow? Or would you check the SVR too? If you would check that, why not use this as the basis of the support for that driver? For a PCI device, you are given basically a 32-bit UID for each device, which is attached to a unique domain, host, bus, device and function number. You manage well here without giving it names.
An Intel processor might return a string for CPUID - mine says it's a "Intel(R) Pentium(R) M 1.70GHz". That string is absolutely useless in determining it's compatibility. It's just a name. It may as well say "Fight Famine In Rwanda" for all the good it does.
I think there are better ways and better places to encode certain properties of the system as a whole (as the system is a lot more than just that single chip) than having a bunch of strings in a property which claim that it is compatible with something else, strictly defined naming conventions and so on across chip ranges. Although we are talking here mostly about two boards with the same chip basically - Lite5200 and Efika, there may be more boards with similar hardware supported, extra hardware supported, new chips which look very much like the 5200 but have slightly different or bugfixed operation (and here is my point) which I do not think you can encode in names and compatibile names.
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Quoted 14 lines · Segher Boessenkool
Quoted 6 lines
(especially a C-based one, OpenBIOS
is 90% Forth which is a terrible lock-in)Isn't the e300 PVR and e300 SVR, or any other device identifier on the chip, a much better differentiator for drivers, than a named compatibility flag?
Only 7 days into the new year, and already we get some great entries for best-joke-of-2007! Too bad they're factually incorrect.
Segher
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 1 line · Matt Sealey
mpc5200b-fec, mpc5200-fec, mpc52xx-fec doesn't tell you anything.
It does. It tells you exactly what you need. It tells you that you are dealing, from the more precise to the more generic:
- It's the implementation of the FEC cell in the mpc5200b
- It's the implementation of the FEC cell in an mpc5200
- it's the implementation of the FEC cell in a mpc52xx family chip
In addition, it really only tells you that it's compatible with those. If we were to have 5500 or 5120, we could just say they are compatible with mpc52xx. It doesn't matter, it's an arbitrary name chosen to represent this cell because it wasn't given us a better name by the constructor. The coutner example is IBM EMAC which has a proper name ... EMAC :-). "FEC" is too generic, freescale calls "FEC" anything that looks remotely like an ethernet controller, so we needed somethign more precise.
Quoted 4 lines
and
then with the 5500 or 5120 or something, you then have worked out you named your original devices wrong, and now need to make it compatible with those?
If they are compatible, they are compatible. In which case, there is nothing wrong in having:
mpc5500-fec\0mpc52xx-fec in the "compatible" property. A bit clumsy maybe but not wrong.
Quoted 1 line
Do you switch features on and off based on a string comparison?
Yes, when the string is well defined. An example is the processor types passed down via AT_PLATFORM to glibc and/or gcc. A string is much better than a number, because at least, it's somewhat self-explanatory.
Quoted 3 lines
What if a certain chip revision has a bug you need to work around (there are plenty in the original MPC5200!), is that meant to be encoded in the "compatible" property too, somehow?
Yes, it should tell you precisely what chip you are talking about, again, from the more specific to the more generic. Of course, it's good to have a more complete information there, thus a "model" property will do no harm (could contain a part number if it's in the SoC bus), that sort of thing in case there is some ambiguity that wasn't foreseen when defining the initial binding, but most of the time, it can be avoided.
Quoted 1 line
Or would you check the SVR too?
Best to avoid. Beside, SVR is completely specific to a given vendor.
Quoted 3 lines
If
you would check that, why not use this as the basis of the support for that driver?
Because the SVR represent the whole chip, every single part/cell in it. We are specifically interested about a specific cell. In many cases, they aren't any significant differences between implementations that would require driver changes for every SVR out there.
Or take the case I gave in another email of Axon. the 4xx coreconnect bits start to be used on non-4xx CPUs as part of a southbridge.... if we had used some SVR-like thingy or even a PVR, suddenly, our entire identification scheme stops working because those exact same cells are no longer tied to a given CPU at all, they are in a south bridge !
Quoted 8 lines
For a PCI device, you are given basically a 32-bit UID for each device, which is attached to a unique domain, host, bus, device and function number. You manage well here without giving it names.
An Intel processor might return a string for CPUID - mine says it's a "Intel(R) Pentium(R) M 1.70GHz". That string is absolutely useless in determining it's compatibility. It's just a name. It may as well say "Fight Famine In Rwanda" for all the good it does.
We don't care, we do care however what family of processor it is, which version of the instruction set it adheres to, and that -is- generally expressed by a string. In addition, we have implementation specific "features" that don't necessarily need to be represented in there. And in case we have to deal with a really specific bug, we -also- want things like part number or mask revision etc... But we don't use that all the time, only for specific quirks.
Quoted 10 lines
I think there are better ways and better places to encode certain properties of the system as a whole (as the system is a lot more than just that single chip) than having a bunch of strings in a property which claim that it is compatible with something else, strictly defined naming conventions and so on across chip ranges. Although we are talking here mostly about two boards with the same chip basically - Lite5200 and Efika, there may be more boards with similar hardware supported, extra hardware supported, new chips which look very much like the 5200 but have slightly different or bugfixed operation (and here is my point) which I do not think you can encode in names and compatibile names.
You logic escapes me.
Ben.
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Matt SealeyThe very fact of matching device to driver by their "name/type/compatible" list
is just the way it's done/specified in OF. Unless I'm deeply mistaken (could always
happen, nobody's perfect heh) ...
The mpc5200b-psc-i2s and mpc5200-psc-i2s are in fact different device ... they
have different register set (compare both datasheet, you'll notice small difference).
They just happen to be handled by the same driver and that driver handles those
difference internally because they're small ...
So back to ethernet since you seem to like it ;) If fsl produce a new chip (let's call
it mpc5321) , either it has _exactly_ the same interface (not a single difference),
they you could call it mpc52xx in your device tree. Sure it looks weird but sorry we can't predict the future and draft things for chips that don't exist yet ...
If it's different, (even slighty, like it handles frame >2048 for example), then
just give it a new name and add it as supported by the same driver. If that's the
5321 entry that matched, then allow frames >2048.
Quoted 1 line
From what I've saw, if they can't even keep the same register set
between two
revision of the same controller, I doubt they will between chips. The I2C controller is a good example, it's almost all the same between several fsl chip,
except for the clocking selection. So as I see it, the I2C driver would have several "compatible" match, for each different variation of the core. How you name those instance is not really important as long as there is no possible
conflicts.
You're proposing (if I understood correctly), that we first match the device to a driver by using the normal name/type/compatible list but with very broad names like "fsl-ethernet", then look at the SVR to know if it's in fact
this driver we need to load (two fsl chip could have completly different ethernet
cores) and also using SVR decide what "quirk" we need to apply. Note I'm not even commenting yet, I first need to be sure of what you're saying,
but the SVR thing will already not work if you're not in a SoC context ...
Regards,
Sylvain
Quoted 40 lines · Matt Sealey
Why is it so funny? I'd love to be enlightened rather than just plain insulted about it.
I don't think any of that can be taken into account by simply giving things *names* and then having a 3-week discussion and committee hearing about how someone gave it a DIFFERENT name. It is even more strange to me, that nobody HAS a standard name for the devices on the MPC5200B, yet we are arguing about what would look nicer in the Linux source code?
Small example:
mpc5200b-fec, mpc5200-fec, mpc52xx-fec doesn't tell you anything, and then with the 5500 or 5120 or something, you then have worked out you named your original devices wrong, and now need to make it compatible with those? Do you switch features on and off based on a string comparison? What if a certain chip revision has a bug you need to work around (there are plenty in the original MPC5200!), is that meant to be encoded in the "compatible" property too, somehow? Or would you check the SVR too? If you would check that, why not use this as the basis of the support for that driver? For a PCI device, you are given basically a 32-bit UID for each device, which is attached to a unique domain, host, bus, device and function number. You manage well here without giving it names.
An Intel processor might return a string for CPUID - mine says it's a "Intel(R) Pentium(R) M 1.70GHz". That string is absolutely useless in determining it's compatibility. It's just a name. It may as well say "Fight Famine In Rwanda" for all the good it does.
I think there are better ways and better places to encode certain properties of the system as a whole (as the system is a lot more than just that single chip) than having a bunch of strings in a property which claim that it is compatible with something else, strictly defined naming conventions and so on across chip ranges. Although we are talking here mostly about two boards with the same chip basically - Lite5200 and Efika, there may be more boards with similar hardware supported, extra hardware supported, new chips which look very much like the 5200 but have slightly different or bugfixed operation (and here is my point) which I do not think you can encode in names and compatibile names.
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 3 lines
The very fact of matching device to driver by their "name/type/compatible" list
is just the way it's done/specified in OF. Unless I'm deeply mistaken
No, you're perfectly right.
Quoted 5 lines
So back to ethernet since you seem to like it ;) If fsl produce a new chip (let's call
it mpc5321) , either it has _exactly_ the same interface (not a single difference),
they you could call it mpc52xx in your device tree.
No, it should be "compatible" = "mpc5321\0mpc52xx", since even if it *supposedly* works identically, there always could be bugs. The actual device version is a huge thing to leave out of the device tree no matter what.
Linux would find it via "mpc52xx", if it has no reason to treat mpc5321 specially, sure. The device tree should still express that information though; it shows information about the hardware to whoever wants to use it (e.g., Linux), it does not *directly* tell Linux how to handle the hardware. This is a good thing.
Quoted 5 lines
If it's different, (even slighty, like it handles frame >2048 for example), then
just give it a new name and add it as supported by the same driver. If that's the
5321 entry that matched, then allow frames >2048.
Yeah.
Segher
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Sylvain MunautQuoted 6 lines · Sylvain Munaut
You're proposing (if I understood correctly), that we first match the device to a driver by using the normal name/type/compatible list but with very broad names like "fsl-ethernet", then look at the SVR to know if it's in fact this driver we need to load (two fsl chip could have completly different ethernet cores) and also using SVR decide what "quirk" we need to apply.
Yes, to a degree. I understand that between MPC5200 and MPC5200B some things did change quite a lot; register places, definitions and so on, even whole ways of supporting a device. I think making a name distinction here and trying to tie it down to some "model number convention" really loses it's shine.
Like Ben said, it's clumsy. Not wrong, but clumsy. I would say the same about the difference between "ac97" and "sound" - it IS a clumsy naming convention and it was contended here, but who really cares as long as the device is detected? Personally I think "sound" is wrong, and ac97 more accurately defines the device (I forget who's side I am on now with that!) according to the OF precedents.
For a PCI device, you do have unique idenitifiers and they are also all present in the device tree for easy scanning (RTAS also provides this) and picking up. Linux is already full of lists of arrays of data in order to attach a device driver to a certain device ID - even with names and compatible nodes, the ethernet and audio and so on drivers for the Efika and Lite5200 have potentially large lists of properties they must check for to determine if the device is correct (one each for mpc5200, mpc5200b, mpc52xx, any compatible future chips will have to list every one of these in their compatible property to 'be correct'..)
So, giving it some other name is no problem at all. Just add the name into the list. You can never get rid of the list, and adding entries.. well.. it may look clumsy, but it is not wrong! Especially if you're the first to name it. Linux should, I am saying, match against what the firmware provides, not vice-versa. I don't think it's logical for firmware vendors to keep changing names or maintain these huge lists... just because of a philosophical difference on 2 lines of code in Linux? Especially when the other OS guys have not made any fuss..
Quoted 2 lines
Note I'm not even commenting yet, I first need to be sure of what you're saying, but the SVR thing will already not work if you're not in a SoC context ...
Yes, I do think that there needs to be a lot less chip naming in there, and some other way of identifying chipset features. I agree with Ben there is some issue when SoC components drop out into other chips, but this fully melds in with my own and bplan's insistance that the "SoC" node idea sucks.
The firmware device tree describes a platform, be that a single board of multiple revisions, or even a range of boards. To place all SoC devices in the root node, we considered clumsy at bplan, so they invented "builtin" instead. This works also for the Pegasos, where rightly there is no SoC. It works for every other board we ported the firmware too under contract to do so (every G4 Freescale board since the 7447, every Marvell dev board, the AMCC "Yosemite"..). It really doesn't matter where you place it, or what you call it in the end, as long as somehow you can determine what kind of device you are looking for.
device_type = "ethernet", well, then it's ethernet. Which kind? Well, it's the kind that can only be available on that particular combination of hardware. If it is an MPC5200B chip or an MPC5200 chip, you can determine exactly which one it is. If it is an MPC8349E or an MPC82xx or something, it may be different, but in general you can determine the exact device you have much easier (one register comparison..) than running through a null-terminated string list, doing comparisons against another list, and you get a wealth of information out of that single 32-bit register value of the SVR that you would not from a simple name. This is where it differs from PCI devices which you can have 10 of them in the same box and need to work out which one came from where, and what esoteric combination the user will use. The device tree is simply there to inform you what is present and what is configured (especially in the case when a unit can be configured as multiple devices and may be dangerous to reconfigure) and on an SoC or even a northbridge or southbridge which implements "non PCI" devices, you have plenty of options for determining exactly what is in them (/builtin/ethernet@1 on Pegasos is the Marvell controller, even though we use a chip with 2 ethernet controllers in it, only one works - there is no PHY card for the other. However rather scarily the option of 1 or 2 ethernet ports in Linux is defined by a kernel config flag... so much for the device tree..)
I am sure there are other things that can be checked and verified against, although there will also be stupid things like.. the only example I can think of is that the MPC7447 and MPC7457 G4 chips have identical PVRs so you can't tell if there are L3 tags or not (or if there is any L3 at the end of it) without some loopy proprietary hardware logic or other system controller feature to inform you. That's not much fun if your CPU board is removable for example :D
Giving it an esoteric device name string, and listing all it's compatible hardware in the firmware tree? Considering Linux maintains it's own little list it checks AGAINST, I see no reason to even name it. It just seems like a waste of time and space where some other heuristic would be a little more specific or more foolproof. If the same driver supports multiple devices, it will check the device to see if it fits with it's list of devices it supports. Like a PCI device driver has a list of compatible vendor/device/function/id inside it. Not just for quirks but just for detection. The compatible property I think is just being misused in the device tree and it's creating more code and more lists and more firmware work to support it.
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Quoted 49 lines
Regards,
Sylvain
Quoted 40 lines · Matt Sealey
Why is it so funny? I'd love to be enlightened rather than just plain insulted about it.
I don't think any of that can be taken into account by simply giving things *names* and then having a 3-week discussion and committee hearing about how someone gave it a DIFFERENT name. It is even more strange to me, that nobody HAS a standard name for the devices on the MPC5200B, yet we are arguing about what would look nicer in the Linux source code?
Small example:
mpc5200b-fec, mpc5200-fec, mpc52xx-fec doesn't tell you anything, and then with the 5500 or 5120 or something, you then have worked out you named your original devices wrong, and now need to make it compatible with those? Do you switch features on and off based on a string comparison? What if a certain chip revision has a bug you need to work around (there are plenty in the original MPC5200!), is that meant to be encoded in the "compatible" property too, somehow? Or would you check the SVR too? If you would check that, why not use this as the basis of the support for that driver? For a PCI device, you are given basically a 32-bit UID for each device, which is attached to a unique domain, host, bus, device and function number. You manage well here without giving it names.
An Intel processor might return a string for CPUID - mine says it's a "Intel(R) Pentium(R) M 1.70GHz". That string is absolutely useless in determining it's compatibility. It's just a name. It may as well say "Fight Famine In Rwanda" for all the good it does.
I think there are better ways and better places to encode certain properties of the system as a whole (as the system is a lot more than just that single chip) than having a bunch of strings in a property which claim that it is compatible with something else, strictly defined naming conventions and so on across chip ranges. Although we are talking here mostly about two boards with the same chip basically - Lite5200 and Efika, there may be more boards with similar hardware supported, extra hardware supported, new chips which look very much like the 5200 but have slightly different or bugfixed operation (and here is my point) which I do not think you can encode in names and compatibile names.
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 10 lines · Matt Sealey
So, giving it some other name is no problem at all. Just add the name into the
list. You can never get rid of the list, and adding entries.. well.. it may
look clumsy, but it is not wrong! Especially if you're the first to name it.
Linux should, I am saying, match against what the firmware provides, not vice-versa. I don't think it's logical for firmware vendors to keep changing
names or maintain these huge lists...
For devices themselves (like serial, fec, ...), if there was only issues there,
I must admit I'd consider just adding the other "compatible" property (look at the current mpc52xx_uart.c driver, it has an entry only for the efika). BUT :
- I'd really prefer to have it with the standard name, we're early
enough to
change it and I love consistency ;)
- The "other" name must _not_ conflict with the official one (and here that's not the case, mpc5200-ata would trigger activation of 5200 errata correction for ATA ... )
- You will anyway release an update so _why_ not change that, I haven't heard so far a single good reason why not ...
- For more "system" parts (like bestcomm/sram/pic), there it's a lot more annoying and I just want the received dt to exactly fit.
Since some changes couldn't be handled by just adding en entry to the of_match
table (chrp type, missing irqs, sram type, ...) , I needed fixups anyway so I
implemented all the fixups I'd like to see and deal with it in a single place,
so at the end I have things exactly like for other 52xx boards ...
Quoted 4 lines
just because of a philosophical difference
on 2 lines of code in Linux? Especially when the other OS guys have not made
any fuss..
But did the other OS support other 52xx board ? (question, I really don't know)
If it only needed another "compatible" entry in the driver list of a couple of
driver, and that there was no way the fw could ever change, I think we would have (reluctantly ;) done it, but the problem is both deeper and the vendor (you ;) have the clear opportunity to fix it better.
Sylvain
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Matt SealeyQuoted 13 lines
device_type= "ethernet", well, then it's ethernet. Which kind? Well, it's the kind that can only be available on that particular combination of hardware. If it is an MPC5200B chip or an MPC5200 chip, you can determine exactly which one it is. If it is an MPC8349E or an MPC82xx or something, it may be different, but in general you can determine the exact device you have much easier (one register comparison..) than running through a null-terminated string list, doing comparisons against another list, and you get a wealth of information out of that single 32-bit register value of the SVR that you would not from a simple name. This is where it differs from PCI devices which you can have 10 of them in the same box and need to work out which one came from where, and what esoteric combination the user will use. The device tree is simply there to inform you what is present and what is configured (especially in the case
Quoted 8 lines
when a unit can be configured as multiple devices and may be dangerous to reconfigure) and on an SoC or even a northbridge or southbridge which implements "non PCI" devices, you have plenty of options for determining exactly what is in them (/builtin/ethernet@1 on Pegasos is the Marvell controller, even though we use a chip with 2 ethernet controllers in it, only one works - there is no PHY card for the other. However rather scarily the option of 1 or 2 ethernet ports in Linux is defined by a kernel config flag... so much for the device tree..)
I don't know where to begin to explain how wrong I think you are ...
To simplify, let's say that a device should be represented by a device node that is pretty much self contained. There should be no need to go look at anything else (processor node, SVR register, whatever). You should be able to uniquely identify every device in the system based on a few properties, and OpenFirmware provides precisely a mecanism to do so.
This mecanism, is the "compatible" property.
If you look closely at the OF PCI binding (since PCI is one of your examples), you'll notice that while pciDDDD,VVVV (device,vendor) is at the beginning of the compatible list, pciCCCC (class) follows.
Again, same logic as explained earlier: you define your device from the more specific to the more generic.
In many cases, a device programming interface is generic (for example, the UHCI, OHCI and EHCI USB host controller drivers, on PCI, match based on the class, not the vendor/device IDs. They only use the later for identifying a handful a chips that need erratas/quirks, they do NOT have and do not need a complete list of all the vendor/device ID combinations for those types of controllers).
That sort of class level matching is also widely used for USB devices.
Thus, the driver -choses- at what level it wants to detect supported devices, based on that, -and- can still explicitely look at the more "precise" indentification when it needs to single out a given revision/version of a piece of hardware.
So this mecanism is great and isn't actually that clumsy at all. What might eventually be clumsy is purely the naming choice when the vendor doesn't provide something nice/unique enough for a given part. For example, in order to properly identify the 52xx "fec" ethernet part, since we don't have a nice/unique codename for it, we use mpc52xx-fec. it's good enough and it's our convention.
For PCI devices, the compatible property is defined to be a fairly long list, which can contain vid/did pairs (and multiple ones as one device cna be made HW compatible with another one) but goes down to class based matching which is more useful for well defined programming interfaces.
Strings are perfect, much better than numbers, as long as we properly define the content of the string.
Ben.
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Segher BoessenkoolQuoted 2 lines · Segher Boessenkool
Quoted 2 lines
(especially a C-based one, OpenBIOS
is 90% Forth which is a terrible lock-in)
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Segher BoessenkoolQuoted 2 lines · Segher Boessenkool
Quoted 2 lines
(especially a C-based one, OpenBIOS
is 90% Forth which is a terrible lock-in)
^^^^^^^
Quoted 3 lines
Only 7 days into the new year, and already we get some great entries for best-joke-of-2007! Too bad they're factually incorrect.
I'll correct that then, FirmWorks (as used on OLPC) is 90% Forth, and I think there is a shortage of Forth coders compared to driver-writing low-level firmware coders who can do C or C++ or any other language for that matter, I think it makes it commercially unusable if you want any kind of turnaround on device support.
OpenBIOS is maybe 30% Forth supplementing a fairly C-based implementation but it is still too much Forth. It also relies on a boot loading stage previous to it (no abstraction!) like UBoot or a PC BIOS to start with, so commercially without some work, it's fairly useless.
Either way, without talking about CodeGen's source release (not commercially usable, and debatably not GPL-compatible either) we're the only company doing Open Firmware for PowerPC devices, with the turnaround time we have (talk to Marvell if you like) and offering the kind of feature support (x86 emulator) we plant on top on request (Linux BSP including working 3D drivers just to prove the firmware works and the emulator is setting up any card they desire correctly).
Maybe IBM's PAPR firmware will be great when it arrives sometime in 2008, and put us out of business, but until then I stand by that :D
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 5 lines
Quoted 2 lines
Quoted 2 lines
(especially a C-based one, OpenBIOS
is 90% Forth which is a terrible lock-in)^^^^^^^
I'll correct that then, FirmWorks (as used on OLPC) is 90% Forth,
Much more than 90%, heh.
Quoted 8 lines
and
I think there is a shortage of Forth coders compared to driver-writing low-level firmware coders who can do C or C++ or any other language for that matter, I think it makes it commercially unusable if you want any kind of turnaround on device support.OpenBIOS is maybe 30% Forth supplementing a fairly C-based implementation but it is still too much Forth.
I really am not going to argue with you about this; you obviously have no experience at all with this stuff, what you're saying is based on preconceptions and nothing more.
Quoted 3 lines
It also relies on a boot loading stage
previous to it (no abstraction!) like UBoot or a PC BIOS to start with, so commercially without some work, it's fairly useless.
It is made to work with LinuxBIOS, instead. Please research your statements.
Quoted 3 lines
Either way, without talking about CodeGen's source release (not commercially usable, and debatably not GPL-compatible either) we're the only company doing Open Firmware for PowerPC devices,
Same comment.
[advertisement snipped here]
It's great you feel so nice about the feature set of your product, but I fail to see what that has to do with your attacks on technical sound ways to do technical stuff.
Anyway, let me explain some very basic things. PowerPC Linux uses an Open Firmware-style device tree to communicate things (mostly what hardware is there) from the firmware or the boot loader to the kernel. You cannot change this. If you really refuse to play within the rules of the game, feel free to create and maintain your own fork of the kernel tree.
Segher
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Matt SealeyQuoted 6 lines
NetBSD, FreeBSD will support the device tree for enumeration and the ports in progress already do, and they also keep the Client Interface open so they have a much better environment to work from than the Linux "cache" of the device tree. If the OF device is there and there is no driver in the OS, at least somehow (slowly) it could be used.
The idea that keeping OF alive is a "much better environment" to work from is at best an illusion.
Quoted 4 lines
#1: As the only Open Firmware platform on this chip, what *we* do matters here, we DO have to offer some stability in operating system support and that means once we fix a device tree it needs to stay fairly similar, which is exactly why we feel justified defining the tree standard for it :D
And that's why we feel justified asking you to change it for the very reason that you did this "definition" behind closed doors, repeately refusing to discuss it publically with the other parties involved in porting linux to the 52xx
Quoted 5 lines
We do have to work with a lot more operating systems and vendors than just Linux, and a "booting without Open Firmware" standard which proves to solve the lack of device tree in U-Boot and other firmwares by making more work for firmware and Linux kernel porting harder, is not really something we are too concerned about.
If that's what you think it is, then you are full of sh*t. Sorry, but I'm getting really badly pissed off by your attitude and that of bPlan. There is a limit to the amount of arrogance I can cope with.
Quoted 7 lines
We hope we offer something a little
better than "recompile your device tree file, embed it into the kernel, hope UBoot is the correct version, or recompile it, and flash and hope for the best". What we offer here is fairly unique; both in that we seem to be the only guys offering Open Firmware (especially a C-based one, OpenBIOS is 90% Forth which is a terrible lock-in), and on the MPC5200 (or PowerPC in general).
Can we cut on the marketing filler a bit here and get back to useful points ?
Quoted 5 lines
We also recognise OF and the device tree is flawed in some areas, RTAS isn't the best thing in the world (but it is totally misused in Linux, in places) so we are not BLIND to the need for change, we just think a few compatibility names and node
device_typechanges are not really where the real problems are.
It's where most of the problems that matter in pratice are. As for "runtime abstraction of hardware", I agree, RTAS sucks. As SMM BIOS sucks. As ACPI tables suck. As keeping OF alive suck badly too. In fact, there isn't a single attempt at doing that sort of thing that has not resulted in creating more problems than it solves, except maybe in some very limited circumstances where the firmware and the OS have been very closely tailored to each other.
In fact, history so far proves that it's a bad idea. It sounds good at first, but when you get down to the details, it doesn't work out.
Get away with it, nothing works as well as a good HW description & documentation and proper drivers for it.
Quoted 4 lines
#2: Personally I think naming device compatibility after the chip they are in is a ridiculous lock-in, also. You may end up with a
device_typeof ethernet, and compatibility flags for 100 SoC chip numbers. This is fairly stupid.
That isn't the rule. The rule for compatible is to name after the functional cell used. However, if it happens that this cell is very specific to a familiy of SOCs (which is common), then go for it. The rule is also that "compatible" is a list from the most specific to the most generic. Thus it _ALSO_ shall contain a name specific to a given implementation in order to allow the driver to differenciate them in order to work around the always present implementation specific errata.
Quoted 4 lines
Isn't the e300 PVR and e300 SVR, or any other device identifier on the chip, a much better differentiator for drivers, than a named compatibility flag? These are present in the device tree (/cpu node?) no? I can't boot mine currently I blew up my power supply unfortunately.
The PVR of the core is probably the worst possible choice there :-) If there's one thing that really doesn't matter much as far as device identification is concerned, it's what core is used on the die.
Quoted 3 lines
I happen to know the next Freescale device won't be called 52xx so we're talking about adding another compatibility naming to it like 51xx or 55xx or 5xxx and changing everyone's firmware and Linux device trees again.
Why ? Who cares what they call the next one ? Once we have a defined binding that properly identify those parts. If their next chip has an FEC that is actually compatible with the 52xx-fec, then it should contain that in the compatible property, whatever the marketing name at the time is. We pick a name that defines a programming interface. It could be anything. It could be the codename of the cell of we knew it. 52xx-fec is as good as anything else, as long as we _agree_ what convention we want to use and keep consistent accross devices. It's an arbitrary choice.
Quoted 4 lines
Or you could switch based on the processor! Cleaner, simpler, and doesn't need each device to have a specially unique name, only a vague "class" without some specific model number in it. If it's an SoC.. this is what the SVR is for!
And totally bogus.
Quoted 3 lines
#3: While the above is basically a case for an SoC node (device parent would have the SVR in it), this does not scale generically or flexibly to non-SoC platforms where devices are nevertheless highly integrated.
? I don't understand your sentence.
Quoted 5 lines
AMD's new processors with integrated CPU and memory controller, will not be classed as SoC's - how would you class them, really? Just a CPU with some devices builtin (aha!) I would say. With the Pegasos, we have the CPU, and the Marvell chip has Ethernet (not on PCI) and SRAM (also just mapped in as memory) but it is not an SoC.
Who cares ? I don't personally think the word "SoC" is indeed very good as, as you pointed out, things can be implemented in different ways with the same parts. However, again, it's just arbitrary naming conventions to represent the specific set of IP blocks those drivers we are writing for.
Take a very good example here: 4xx and cell.... The 4xx processors use the IBM CoreConnect library of parts for their on-chip devices (EMAC ethernet, MAL, PLB5, PLB4, OPB busses, etc...). They are technically SoCs. However, we now have this southbridge for the Cell processor which happens to ... contain a PLB5 and all ther other bits & pieces. In fact, it also contains a 405 so it could be considered a SoC... or not :-) The point is:
IT DOES NOT MATTER !
It's all a naming convention. We call plb5 ... plb5 heh ! The emac will have a compatible property with something like "emac-axon\0emac4\0emac". That is, from the more specific to the more generic: The axon implementation of emac, which is an emac4 cell which is an emac familiy of ethernet cores. This is a convention. Once decided (there will be a public discussion, unless nobody disagrees with my proposal there), if tomorrow, a version of AMD64 is released with an EMAC in it and x86 folks want an OF-like device-tree (which is something that might be happening sooner than expected), it will hopefully follow that convention.
And this has nothing to do with booting without OF. In fact, we should split that document between what is indeed booting without OF (the format of the flattened device-tree that can be provided by zImage or uboot if you don't have a real OF) from what is in practice OF bindings for a variety of chips out there.
Quoted 5 lines
We just don't want to clutter up the root node with devices, otherwise it would basically be more of a device shrubbery. The e300 SVR is also defined in the manual (as is the e500 equivalent, and e200 equivalent) as a CPU core feature, not a SoC-device feature, so this kind of differentiation and numbering belongs in the CPU node.
If there is something that does NOT belong in the CPU nodes, it's anything related to the on chip devices. Bang bang... try again.
Quoted 5 lines
~
Those are MY problems with the device tree specification you have with "Booting Without Open Firmware". I am sure the other guys have other things to say.
You don't seem to understand it much though.
Ben.
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Matt SealeyQuoted 6 lines · Matt Sealey
I would expect that it is so much easier, even if it does not strike you as too "clean", to forget about naming compatibility nodes here, and concentrate on the real showstoppers. I would consider the AC97 PSC not being set up, the IrDA PSC not being set up, the disk numbering and any problems we're having with USB devices (argh) to be real showstoppers.
I'm inclined to agree. What I said was "since they have showstopper bugs to fix, we might as well let them fix the less important device-tree stuff at the same time."
Please can we have a firmware with the important stuff fixed, and preferably the simple and uncontentious device-tree fixes too; like perhaps no longer pretending to be CHRP?
The other device-tree stuff you're arguing about is less important, because it isn't _that_ hard to fix it up in prom_init() -- although it would have been nice if the debate had happened when people were trying to put the spec together in the first place, rather than only now.
Signature · 1 line
dwmw2
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 11 lines · David Woodhouse
Quoted 6 lines · Matt Sealey
I would expect that it is so much easier, even if it does not strike you as too "clean", to forget about naming compatibility nodes here, and concentrate on the real showstoppers. I would consider the AC97 PSC not being set up, the IrDA PSC not being set up, the disk numbering and any problems we're having with USB devices (argh) to be real showstoppers.
I'm inclined to agree. What I said was "since they have showstopper bugs to fix, we might as well let them fix the less important device-tree stuff at the same time."
The problem is that the firmware development and testing cycle is pretty intensive; if we change any string in the code, if it is publically present then it needs to be passed through a set of tests to make sure it is working. If it means collaborating with Linux developers it means sending out firmware updates, getting results back.. just for changing one string to another.
When it doesn't make any difference at all ("sram" and "memory" and "ac97" to "sound", the difference between mpc5200-blah and mpc5200b-blah and a couple of stray null characters) I feel there's no point whatsoever in doing it when it entails that much work.
In the Pegasos 1.2 firmware we definitely had some stupid spelling mistakes (one RTAS token at LEAST was wrong :) but since nobody ever used them, they were never really fixed. The CPU node presented l3 cache data for the 7447 even though we never made a 7457 card and could differentiate between them at the point that firmware was out.
But nobody cares. It doesn't make a single bit of functional difference where the thing is fixed if it IS needed though. It may be cleaner in the firmware, but what I saw here is Sylvain saying that these fixes will "never be accepted in the mainline" because they should be fixed in firmware. We can't fix the 1.3 firmware, it's already on the first production batch, end of story. No committee or discussion does anything to change that. We can and will fix as much as is relevant in the next one but if we can't.. does that mean Linux is intentionally crippling itself?
I don't care if it makes the platform look full of bugs as long as Linux runs on it. I am concerned when a real hardware bug is marked as a platform bug (Via IDE IRQ quirk for Pegasos is a good example, this is a Via chipset bug, not a PowerPC or Pegasos-specific thing) but these are really just platform bugs, that we would like to see code fixed up for it. If it means MPC5200 drivers have Efika fixes in it, then *sorry* but I am thinking of people who want to use PowerPC Linux here.
I really want to give the customers the choice; because sometime soon we may have NetBSD, FreeBSD, QNX, whatever else on the board, and we could be saying "well, these OS's all work fine, Linux does too up to a point, but they have crippled it" because it would be the truth.
Quoted 3 lines
Please can we have a firmware with the important stuff fixed, and preferably the simple and uncontentious device-tree fixes too; like perhaps no longer pretending to be CHRP?
That should be most definitely fixed in February. We like CHRP for what it is but the reality of it is that it's so weakly defined in Linux and not used at all in any other OS, there is no point to it. RTAS is totally misused in places. We have a solution for it.. maybe a little contentious but at least it will not be presenting our systems as a standard they may only half-be (Pegasos was VERY CHRP besides the interrupt controller, Efika is about as CHRP as my grandmother's handbag).
Quoted 4 lines
The other device-tree stuff you're arguing about is less important, because it isn't _that_ hard to fix it up in
prom_init()-- although it would have been nice if the debate had happened when people were trying to put the spec together in the first place, rather than only now.
We had the final firmware pretty much completed and in a no-minor-tweaks phase before any spec was ever considered on this mailing list, which is the whole problem. Being a little pioneering about it means we fell afoul of the guys who think everything should be discussed to death before doing it. Efika was already delayed 6 months because of errata and RoHS problems, and we just don't have the time in any case to sit here and discuss a million minor things when the number of boards in production was a big fat zero, or now that boards are out in the field and need support.
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Re: [PATCH] Probe Efika platform before CHRP.
Ok, let's try to at least be a little productive:
Could you tell us at least what in that list will be taken care of, what you might consider and what you really will never change while you're standing ;)
Her we go :
1) What I think should absolutely be fixed : (non compliancy with the ieee1275 document are marked as bug) :
- PSC init (for PSC2 and PSC6 you said as well): Theses are obviously bugs, no argument there.
- Missing IRQ for the sound node : Again, this is obviously missing, the interrupt does exists and it should be in the dt, whether or not you think the driver will use it or not.
- Partition numbering problem (again, an obvious bug)
- Double \0 : This one is a bug (just confirmed it re-reading the specs) Altough I must admit the kernel just won't care it's nevertheless a bug and should be
- "chrp" type : If I understood correctly your previous mail you're ok with it.
- "memory" type of the SRAM: I quote the 1275.pdf, section 3.7.6 : "In this context, 'memory' refers to traditional RAM, suitable for temporary storage of data". I think we can agree sram is not conventionnal RAM. You're free (and should) still use the reg and available property but not the type ram. I must also say that the current fixup I implemented doesn't really work because this can only be fixed too late ... (node already processed and added as normal memory).
2) What I'd really like to see fixed (but that I can't mark as 'bugs' as obviously as the other)
- The compatible properties of system node (like sram, bestcomm and pic) should at least include the mpc52xx-... strings as defined in the 'linux' bindings. Better than include, be exacty what has been defined but include would "do the trick".
3) 'real' Device compatible names : Those can be handled in the driver, eventually differentiating from other using the name ... (Although the ata node is really boring because it _perfectly_ match the defined binding of the 5200 ATA ...)
Without (1), we would need a nvramrc (to fix very early stuff), and since it's needed, might as well fix everything in nvramrc. With only (1), we will still need a fixup, but that could be done in efika_fixup in prom_init.c, since we need fixup there, we will probably do everything there.
With (1) & (2) the efika will boot, just some driver won't work so we need to add some more entry in their of_match but that's acceptable.
Quoted 4 lines
When it doesn't make any difference at all ("sram" and "memory" and "ac97" to "sound", the difference between mpc5200-blah and mpc5200b-blah and a couple of stray null characters) I feel there's no point whatsoever in doing it when it entails that much work.
Well, you're gonna need to change _some_ strings at least (if not, I'm just loosing my time here ...) I can hardly believe changing a string to another is that hard, especially if it's to comply to the specification you claim to be compatible with ;p
(and 'ac97' should be 'sound' as it is now, certainly don't go change it ;)
Quoted 2 lines
We can't fix the 1.3 firmware, it's already on the first production batch, end of story.
You mean the new firmware can't be flashed to the old boards ?
Sylvain
PS: I know there are already fixup but ... my mistake I keep dreaming of a world when things do get fixed without hacks ;-)
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 8 lines · Sylvain Munaut
2) What I'd really like to see fixed (but that I can't mark as 'bugs' as obviously as the other)
- The compatible properties of system node (like sram, bestcomm and pic) should at least include the mpc52xx-... strings as defined in the 'linux' bindings. Better than include, be exacty what has been defined but include would "do the trick".
Forgot something here :
The "bestcomm" node should have all the interrupts for all task in it's interrupts property instead of a non existent IRQ. Currently it does have <2 0 3> which doesn't really exists (it's the casaded interrupt IIRC, this is from memory, no time to check that now).
And it should have <3 0 0> <3 1 0> .... <3 15 0>.
Because those are really bestcomm interrupt and I think they do belong there.
I know I can find thos in your "bestcomm-txtask bestcomm-rxtask" nodes for preloaded tasks and it's good they are there. But what if I want to use DMA for PSC6 where you didn't load anytask ... I can find some free ram in the sram node, some non-used tasks id but I got nowhere to find the interrupt ... So I'd like to have them there as well ...
Sylvain
Re: [PATCH] Probe Efika platform before CHRP.
This will definitely change in the next firmware from the discussion I've had with Gerald.
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Quoted 30 lines · Sylvain Munaut
Quoted 8 lines · Sylvain Munaut
2) What I'd really like to see fixed (but that I can't mark as 'bugs' as obviously as the other)
- The compatible properties of system node (like sram, bestcomm and pic) should at least include the mpc52xx-... strings as defined in the 'linux' bindings. Better than include, be exacty what has been defined but include would "do the trick".Forgot something here :
The "bestcomm" node should have all the interrupts for all task in it's interrupts property instead of a non existent IRQ. Currently it does have <2 0 3> which doesn't really exists (it's the casaded interrupt IIRC, this is from memory, no time to check that now).
And it should have <3 0 0> <3 1 0> .... <3 15 0>.
Because those are really bestcomm interrupt and I think they do belong there.I know I can find thos in your "bestcomm-txtask bestcomm-rxtask" nodes for preloaded tasks and it's good they are there. But what if I want to use DMA for PSC6 where you didn't load anytask ... I can find some free ram in the sram node, some non-used tasks id but I got nowhere to find the interrupt ... So I'd like to have them there as well ...
Sylvain
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Sylvain MunautQuoted 14 lines · Sylvain Munaut
1) What I think should absolutely be fixed : (non compliancy with the ieee1275 document are marked as bug) :
- PSC init (for PSC2 and PSC6 you said as well): Theses are obviously bugs, no argument there.
- Missing IRQ for the sound node : Again, this is obviously missing, the interrupt does exists and it should be in the dt, whether or not you think the driver will use it or not.
- Partition numbering problem (again, an obvious bug)
- Double \0 : This one is a bug (just confirmed it re-reading the specs) Altough I must admit the kernel just won't care it's nevertheless a bug and should be
- "chrp" type : If I understood correctly your previous mail you're ok with it.
- "memory" type of the SRAM: I quote the 1275.pdf, section
Quoted 9 lines
3.7.6 : "In this context, 'memory' refers to traditional RAM, suitable for temporary storage of data". I think we can agree sram is not conventionnal RAM. You're free (and should) still use the reg and available property but not the type ram. I must also say that the current fixup I implemented doesn't really work because this can only be fixed too late ... (node already processed and added as normal memory).
Oops, forgot something here as well :
Check for typo, for example "#interuppt-cells" vs "#interrupt-cells"
Sylvain
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 3 lines · Sylvain Munaut
Quoted 1 line
Oops, forgot something here as well :
Check for typo, for example "#interuppt-cells" vs "#interrupt-cells"
See what I mean?
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Re: [PATCH] Probe Efika platform before CHRP.
In reply to Sylvain MunautQuoted 6 lines · Sylvain Munaut
Ok, let's try to at least be a little productive:
Quoted 2 lines
We can't fix the 1.3 firmware, it's already on the first production batch, end of story.
You mean the new firmware can't be flashed to the old boards ?
I mean we don't have the new firmware out and there is two months of Linux usage to go before it will be.. at least.. granted we will be having this discussion of what will need to be changed or not at that time, yet again, because of some other changes we make.
However, changing a string IS a big deal when it needs to be tested and quality assured for a commercial release. If it makes no difference at all, it is a lot easier and less time consuming simply not to change it :D
Signature · 2 lines
Matt Sealey <[email protected]> Genesi, Manager, Developer Relations
Re: [PATCH] Probe Efika platform before CHRP.
Quoted 4 lines · Matt Sealey
I mean we don't have the new firmware out and there is two months of Linux usage to go before it will be.. at least.. granted we will be having this discussion of what will need to be changed or not at that time, yet again, because of some other changes we make.
Or maybe you could tell us now what you're planning to change/add ...
If not, that will be exactly the same thing, if for whatever reason we need more infos in the device tree and that's not currently specified, we'll define some binding for what's missing that will end up being incompatible with what you did during 2 month in secret and in the mean time the whole driver/support code will have been written based on what has been discussed in public ...
Quoted 2 lines
This will definitely change in the next firmware from the discussion I've had with Gerald.
Could you be a little more precise about what you're planning on 'fixing' from the list I posted ? (like point 1, 2 but not 3. Or 1a, 1b 2a & 2b ...) ...
Sylvain
PATCH 1/1 powerpc export rtas_set_slot_reset()
In reply to Paul MackerrasHi Paul,
A patch that provides an export for a bug being worked on the IPR evice driver.
--linas
Export the symbol for the PCI slot reset function; this will allow it to be called from a module.
----
arch/powerpc/platforms/pseries/eeh.c | 1 + 1 file changed, 1 insertion(+)
arch/powerpc/platforms/pseries/eeh.c +1 -0
@@ -626,6 +626,7 @@ int rtas_set_slot_reset(struct pci_dn *preturn rc; } +EXPORT_SYMBOL_GPL (rtas_set_slot_reset);/* ------------------------------------------------------- */ /** Save and restore of PCI BARs
Re: [PATCH] powerpc export rtas_set_slot_reset()
Quoted 3 lines · Linas Vepstas
Hi Paul,
A patch that provides an export for a bug being worked on the IPR evice driver.
This is not the API that I tested with... Here is the patch I was using in my test. This is not ready to apply as we are still working issues with various hardware...
Brian
Signature · 3 lines
Brian King eServer Storage I/O IBM Linux Technology Center
The following patch adds a generic API to assert and deassert PCI reset to a device on a pSeries machine. This is needed to solve an issue with a new ipr PCI-E adapter which causes EEH errors when running BIST. The only way to reset this adapter without causing an EEH error is to use PCI reset.
---
linux-2.6-bjking1/arch/powerpc/platforms/pseries/eeh.c | 42 +++++++++++++++++ linux-2.6-bjking1/include/asm-powerpc/ppc-pci.h | 3 + 2 files changed, 45 insertions(+)
arch/powerpc/platforms/pseries/eeh.c renamed from arch/powerpc/platforms/pseries/eeh.c~powerpc_slot_reset_api +42 -0
@@ -591,6 +591,48 @@ static void __rtas_set_slot_reset(structmsleep (PCI_BUS_SETTLE_TIME_MSEC ); } +/** +* pci_set_slot_reset - Assert PCI reset to the PCI slot +* @dev: pci device struct +* +* After asserting PCI reset, the caller should wait for +* 100 milliseconds or longer. +* +* Return value: +* 0 if success +**/ +int pci_set_slot_reset (struct pci_dev *dev) +{ +struct device_node *dn =pci_device_to_OF_node (dev); +struct pci_dn *pdn =PCI_DN (dn); + +rtas_pci_slot_reset (pdn,1 ); +return 0 ; +} +EXPORT_SYMBOL_GPL (pci_set_slot_reset); + +/** +* pci_clear_slot_reset - Clear PCI reset to the PCI slot +* @dev: pci device struct +* +* After clearing PCI reset, the caller should wait 1.8 seconds +* or longer for the bus to stabilize and the device to come +* ready. +* +* Return value: +* 0 if success +**/ +int pci_clear_slot_reset (struct pci_dev *dev) +{ +struct device_node *dn =pci_device_to_OF_node (dev); +struct pci_dn *pdn =PCI_DN (dn); + +eeh_clear_slot (pdn->node,EEH_MODE_ISOLATED ); +rtas_pci_slot_reset (pdn,0 ); +return 0 ; +} +EXPORT_SYMBOL_GPL (pci_clear_slot_reset); +int rtas_set_slot_reset (struct pci_dn *pdn) {int i, rc;
include/asm-powerpc/ppc-pci.h renamed from include/asm-powerpc/ppc-pci.h~powerpc_slot_reset_api +3 -0
@@ -124,6 +124,9 @@ void eeh_clear_slot (struct device_node/* Find the associated "Partiationable Endpoint" PE */ struct device_node *find_device_pe (struct device_node *dn); +int pci_set_slot_reset (struct pci_dev *dev); +int pci_clear_slot_reset (struct pci_dev *dev); +#endif #endif /* __KERNEL__ */
_
PATCH 1/1 powerpc fixup error message
In reply to Paul MackerrasPaul,
Please apply this somewhat trite patch; it did come up in conversations.
Heads up:
I'm debugging a more serious problem for power4, concerning multi-function adapters & eeh recovery, and would like to get that patch in if/when I get it figured out and fixed.
--linas
Clarify error message re EEH permanent failure.
----
arch/powerpc/platforms/pseries/eeh_driver.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
arch/powerpc/platforms/pseries/eeh_driver.c +2 -1
@@ -446,7 +446,8 @@ excess_failures:*/ printk (KERN_ERR "EEH: PCI device at location=%s driver=%s pci addr=%s \n" -"has failed %d times and has been permanently disabled. \n" +"has failed %d times in the last hour " +"and has been permanently disabled. \n" "Please try reseating this device or replacing it.\n" , location, drv_str, pci_str, frozen_pdn->eeh_freeze_count); goto perm_error;
PATCH 1/1 powerpc initialize pci device channel state.
In reply to Paul MackerrasPaul,
This patch is critical. Even if the last few looked to be of low importance, this one is not.
--linas
Initialize the pci device pci channel state. This is critical for having the pci_channel_offline() routine (in pci.h) to function correctly.
----
arch/powerpc/kernel/pci_64.c | 1 + 1 file changed, 1 insertion(+)
arch/powerpc/kernel/pci_64.c +1 -0
@@ -360,6 +360,7 @@ struct pci_dev *of_create_pci_dev(structDBG (" class: 0x%x\n" , dev->class); dev->current_state =4 ;/* unknown power state */ + dev->error_state = pci_channel_io_normal;if (!strcmp (type,"pci" ) || !strcmp (type,"pciex" )) {/* a PCI-PCI bridge */