net: appletalk: remove Apple/Farallon LocalTalk PC support
Jakub Kicinski
maintainerPATCH 1/1 net: appletalk: remove Apple/Farallon LocalTalk PC support
Looks like all the changes to this driver had been tree-wide refactoring since git era begun. The driver is using virt_to_bus() we should make it use more modern DMA APIs but since it's unlikely to be getting any use these days delete it instead. We can always revert to bring it back.
---
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
--- .../device_drivers/appletalk/index.rst | 1 - .../device_drivers/appletalk/ltpc.rst | 144 -- drivers/net/appletalk/Kconfig | 11 - drivers/net/appletalk/Makefile | 1 - drivers/net/appletalk/ltpc.c | 1277 ----------------- drivers/net/appletalk/ltpc.h | 74 - 6 files changed, 1508 deletions(-) delete mode 100644 Documentation/networking/device_drivers/appletalk/ltpc.rst delete mode 100644 drivers/net/appletalk/ltpc.c delete mode 100644 drivers/net/appletalk/ltpc.h
Documentation/networking/device_drivers/appletalk/index.rst +0 -1
@@ -9,7 +9,6 @@ AppleTalk Device Drivers :maxdepth: 2 cops - ltpc .. only:: subproject and html
Documentation/networking/device_drivers/appletalk/ltpc.rst deleted +0 -144
144 lines not on this page · raw message
drivers/net/appletalk/Kconfig +0 -11
@@ -39,17 +39,6 @@ config DEV_APPLETALK connect to the AppleTalk network, say Y. -config LTPC - tristate "Apple/Farallon LocalTalk PC support" - depends on DEV_APPLETALK && (ISA || EISA) && ISA_DMA_API && VIRT_TO_BUS - help - This allows you to use the AppleTalk PC card to connect to LocalTalk - networks. The card is also known as the Farallon PhoneNet PC card. - If you are in doubt, this card is the one with the 65C02 chip on it. - You also need version 1.3.3 or later of the netatalk package. - This driver is experimental, which means that it may not work. - See the file <file:Documentation/networking/device_drivers/appletalk/ltpc.rst>. - config COPS tristate "COPS LocalTalk PC support" depends on DEV_APPLETALK && ISA
drivers/net/appletalk/Makefile +0 -1
@@ -5,4 +5,3 @@ obj-$(CONFIG_IPDDP) += ipddp.o obj-$(CONFIG_COPS) += cops.o -obj-$(CONFIG_LTPC) += ltpc.o
drivers/net/appletalk/ltpc.c deleted +0 -1277
1,277 lines not on this page · raw message
drivers/net/appletalk/ltpc.h deleted +0 -74
74 lines not on this page · raw message
Signature · 1 line
2.34.3
Quoted 8 lines · Jakub Kicinski
Looks like all the changes to this driver had been tree-wide refactoring since git era begun. The driver is using
virt_to_bus()we should make it use more modern DMA APIs but since it's unlikely to be getting any use these days delete it instead. We can always revert to bring it back.Signed-off-by Jakub Kicinski <[email protected]>
Removing this driver sounds good to me, your description makes sense and it gets us closer to completely removing virt_to_bus() in the future.
I think however, if we remove this driver, we need to discuss removing the last remaining localtalk driver (CONFIG_COPS) and possibly the localtalk bits in net/appletalk along with it.
Doug Brown suggested removing COPS last year for entirely different reasons[1] but never got a reply. I suppose that is a sign that nobody cared about the driver enough, but we should remove it. He also mentioned working on a new localtalk driver, though I don't think he posted that one yet.
Arnd
Jakub Kicinski
maintainerQuoted 25 lines · On Mon, 9 May 2022 19:14:42 +0200 Arnd Bergmann wrote:
Quoted 8 lines · Jakub Kicinski
Looks like all the changes to this driver had been tree-wide refactoring since git era begun. The driver is using
virt_to_bus()we should make it use more modern DMA APIs but since it's unlikely to be getting any use these days delete it instead. We can always revert to bring it back.Signed-off-by Jakub Kicinski <[email protected]>Removing this driver sounds good to me, your description makes sense and it gets us closer to completely removing
virt_to_bus()in the future.Acked-by Arnd Bergmann <[email protected]>I think however, if we remove this driver, we need to discuss removing the last remaining localtalk driver (
CONFIG_COPS) and possibly the localtalk bits in net/appletalk along with it.Doug Brown suggested removing COPS last year for entirely different reasons[1] but never got a reply. I suppose that is a sign that nobody cared about the driver enough, but we should remove it. He also mentioned working on a new localtalk driver, though I don't think he posted that one yet.
Quoted 3 lines
Arnd
Very interesting, thanks!
Removing COPS and appletalk makes perfect sense to me (minus what Doug has plans to use, obviously).
I'm taking notes for "next steps" while trying to strategically cast a narrow net, hitting VIRT_TO_BUS only now.
Quoted 6 lines · Jakub Kicinski
Quoted 3 lines · On Mon, 9 May 2022 19:14:42 +0200 Arnd Bergmann wrote:
I think however, if we remove this driver, we need to discuss removing the last remaining localtalk driver (
CONFIG_COPS) and possibly the localtalk bits in net/appletalk along with it.Removing COPS and appletalk makes perfect sense to me (minus what Doug has plans to use, obviously).
I also think removing the COPS driver is a great idea. I actually ended up buying a compatible card in the hopes of working on that driver to change it to load the firmware through the firmware API, but the licensing situation with the firmware blobs kind of brought that idea to a standstill. I would be very surprised if anybody is actually using LocalTalk ISA cards these days anyway, so it's probably not worth the effort to maintain it.
There have been a few "modern" LocalTalk interface projects. One is mine, which I haven't found time to finish, but I was able to get working in the kernel with a lt0 network interface. I suspect I was the only one in the last decade to actually use the LocalTalk code in modern kernel versions, because it was crashing until I fixed a bug involving too short of a header length being allocated. There's another more recent LocalTalk project called TashTalk [1]. A kernel driver could be developed for it using serdev or a tty ldisc, but all of the current development seems focused on the userspace side.
With that in mind, I personally wouldn't be sad to see the entire LocalTalk interface support stripped from the kernel, as long as EtherTalk support can remain. There is still a decent sized community of users who are using it to talk with classic Macs using netatalk 2.x. So most of the stuff in net/appletalk is still relevant today for us.
Might as well remove CONFIG_IPDDP too. It actually -interferes- with the current way that people do MacIP gateways through userspace with macipgw [2]. I'm not aware of anyone actually using the kernel's implementation.
Doug
[1] https://github.com/lampmerchant/tashtalk
[2] https://github.com/jasonking3/macipgw#kernel
Quoted 33 lines · Doug Brown
Quoted 6 lines · Jakub Kicinski
Quoted 3 lines · On Mon, 9 May 2022 19:14:42 +0200 Arnd Bergmann wrote:
I think however, if we remove this driver, we need to discuss removing the last remaining localtalk driver (
CONFIG_COPS) and possibly the localtalk bits in net/appletalk along with it.Removing COPS and appletalk makes perfect sense to me (minus what Doug has plans to use, obviously).
I also think removing the COPS driver is a great idea. I actually ended up buying a compatible card in the hopes of working on that driver to change it to load the firmware through the firmware API, but the licensing situation with the firmware blobs kind of brought that idea to a standstill. I would be very surprised if anybody is actually using LocalTalk ISA cards these days anyway, so it's probably not worth the effort to maintain it.
There have been a few "modern" LocalTalk interface projects. One is mine, which I haven't found time to finish, but I was able to get working in the kernel with a lt0 network interface. I suspect I was the only one in the last decade to actually use the LocalTalk code in modern kernel versions, because it was crashing until I fixed a bug involving too short of a header length being allocated. There's another more recent LocalTalk project called TashTalk [1]. A kernel driver could be developed for it using serdev or a tty ldisc, but all of the current development seems focused on the userspace side.
With that in mind, I personally wouldn't be sad to see the entire LocalTalk interface support stripped from the kernel, as long as EtherTalk support can remain. There is still a decent sized community of users who are using it to talk with classic Macs using netatalk 2.x. So most of the stuff in net/appletalk is still relevant today for us.
Quoted 3 lines
Might as well remove
CONFIG_IPDDPtoo. It actually -interferes- with the current way that people do MacIP gateways through userspace with macipgw [2]. I'm not aware of anyone actually using the kernel's implementation.
Thanks for all the background information!
If I understand this correct, this means we could remove all of drivers/net/appletalk/ except for the CONFIG_ATALK Kconfig entry, and also remove net/appletalk/dev.c and a few bits of net/appletalk that reference localtalk device structures and their ioctls, right?
What about appletalk over PPP (phase1 probing in aarp.c) and ARPHRD_LOCALTLK support in drivers/net/tun.c? Are these still useful without localtalk device support?
Arnd
Quoted 4 lines · Arnd Bergmann
If I understand this correct, this means we could remove all of
drivers/net/appletalk/except for theCONFIG_ATALKKconfigentry, and also removenet/appletalk/dev.cand a few bits of net/appletalk that reference localtalk device structures and their ioctls, right?
Yes, I believe so. At that point, would Kconfig get moved to net/appletalk instead? (Just wondering out of my own curiosity!)
Quoted 3 lines
What about appletalk over PPP (phase1 probing in
aarp.c) andARPHRD_LOCALTLKsupport indrivers/net/tun.c? Are these still useful without localtalk device support?
I don't feel qualified enough to answer those ones definitively, but it looks to me like the ARPHRD_LOCALTLK support in net/tun.c could be stripped out, because tun_get_addr_len only gets called on a struct net_device's type, and stripping out LocalTalk would make that condition impossible (I think?)
The AppleTalk over PPP stuff probably allows Linux to be an AppleTalk Remote Access server. I'm not aware of anyone using that capability, (or if it even still works) but I would consider it distinct from LocalTalk.
I would definitely be happy to test any patches to make sure that EtherTalk still works with netatalk afterward!
Doug
Quoted 12 lines · Doug Brown
Quoted 4 lines · Arnd Bergmann
If I understand this correct, this means we could remove all of
drivers/net/appletalk/except for theCONFIG_ATALKKconfigentry, and also removenet/appletalk/dev.cand a few bits of net/appletalk that reference localtalk device structures and their ioctls, right?Yes, I believe so. At that point, would
Kconfigget moved to net/appletalk instead? (Just wondering out of my own curiosity!)Quoted 3 lines
What about appletalk over PPP (phase1 probing in
aarp.c) andARPHRD_LOCALTLKsupport indrivers/net/tun.c? Are these still useful without localtalk device support?
Quoted 5 lines
I don't feel qualified enough to answer those ones definitively, but it looks to me like the
ARPHRD_LOCALTLKsupport innet/tun.ccould be stripped out, becausetun_get_addr_lenonly gets called on a struct net_device's type, and stripping out LocalTalk would make that condition impossible (I think?)
Right, I came to the same conclusion here.
Quoted 3 lines
The AppleTalk over PPP stuff probably allows Linux to be an AppleTalk Remote Access server. I'm not aware of anyone using that capability, (or if it even still works) but I would consider it distinct from LocalTalk.
I dug around in the early git history for this one, but I'm also not sure if this is meant to still work. I see that PPPTALK support was added to net/appletalk by Alan Cox in linux-1.3.78 (1996), based on the localtalk support, and it continues to exist there along ethertalk and localtalk.
I also looked at the git history for the pppd user space, and I find no indication of appletalk ever being supported there, this all looks IPv4/IPv6 specific. There was support for PPP_IPX until it was dropped this year (the kernel side got removed in 2018), but never for PPP_AT.
Adding Paul Mackerras to Cc, he might know more about it.
Quoted 2 lines
I would definitely be happy to test any patches to make sure that EtherTalk still works with netatalk afterward!
Arnd
This patch was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:
Hello:
This patch was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:
Quoted 9 lines · On Mon, 9 May 2022 08:01:30 -0700 you wrote:
Looks like all the changes to this driver had been tree-wide refactoring since git era begun. The driver is using
virt_to_bus()we should make it use more modern DMA APIs but since it's unlikely to be getting any use these days delete it instead. We can always revert to bring it back.Signed-off-by Jakub Kicinski <[email protected]>[...]
Here is the summary with links:
- [net-next] net: appletalk: remove Apple/Farallon LocalTalk PC support
https://git.kernel.org/netdev/net-next/c/03dcb90dbf62
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
Quoted 5 lines · Arnd Bergmann
indication of appletalk ever being supported there, this all looks IPv4/IPv6 specific. There was support for
PPP_IPXuntil it was dropped this year (the kernel side got removed in 2018), but never forPPP_AT.
Adding Paul Mackerras to Cc, he might know more about it.
I waited a bit before chipping in, as I think Paul would know more.
The ATCP stuff was in at least a few vendor branches, but I don't think it ever made it into the main distribution. These commits seem to be where the (disabled by default) references to it first appeared:
commit 50c9469f0f683c7bf8ebad9b7f97bfc03c6a4122
Author: Paul Mackerras <[email protected]> Date: Tue Mar 4 03:32:37 1997 +0000
add defs for appletalk
commit 01548ef15e0f41f9f6af33860fb459a7f578f004
Author: Paul Mackerras <[email protected]> Date: Tue Mar 4 03:41:17 1997 +0000
connect time stuff gone to auth.c,
don't die on EINTR from opening tty,
ignore NCP packets during authentication,
fix recursive signal problem in kill_my_pg
The disabled-by-default parts were likely support contributions for those other distributions. (Very likely in BSD.)
I would've thought AppleTalk was completely gone by now, and I certainly would not be sad to see the dregs removed from pppd, but there was a patch release on the netatalk package just last month, so what do I know?
(The only possible reason I can see to keep any ATCP bits around at all is to make sure we can write nice-looking log messages -- to say we're rejecting "AppleTalk Control Protocol" rather than "unknown 8029." But that'd be a very minor feature.)
Signature · 1 line
James Carlson 42.703N 71.076W FN42lq08 <[email protected]>
Quoted 20 lines · James Carlson
Quoted 5 lines · Arnd Bergmann
indication of appletalk ever being supported there, this all looks IPv4/IPv6 specific. There was support for
PPP_IPXuntil it was dropped this year (the kernel side got removed in 2018), but never forPPP_AT.
Adding Paul Mackerras to Cc, he might know more about it.I waited a bit before chipping in, as I think Paul would know more.
The ATCP stuff was in at least a few vendor branches, but I don't think it ever made it into the main distribution. These commits seem to be where the (disabled by default) references to it first appeared:
commit
50c9469f0f683c7bf8ebad9b7f97bfc03c6a4122Author: Paul Mackerras <[email protected]> Date: Tue Mar 4 03:32:37 1997 +0000add defs for appletalk
Quoted 8 lines
commit
01548ef15e0f41f9f6af33860fb459a7f578f004Author: Paul Mackerras <[email protected]> Date: Tue Mar 4 03:41:17 1997 +0000connect time stuff gone to
auth.c,
don't die on EINTR from opening tty,
ignore NCP packets during authentication,
fix recursive signal problem inkill_my_pg
Right, I had seen those in the git history, but neither of them actually does anything with appletak.
Quoted 6 lines
The disabled-by-default parts were likely support contributions for those other distributions. (Very likely in BSD.)
I would've thought AppleTalk was completely gone by now, and I certainly would not be sad to see the dregs removed from pppd, but there was a patch release on the netatalk package just last month, so what do I know?
I think netatalk 3.0 dropped all appletalk protocol stuff a long time ago and only supports AFP over IP.
Arnd
Quoted 8 lines · Arnd Bergmann
Quoted 4 lines · James Carlson
I would've thought AppleTalk was completely gone by now, and I certainly would not be sad to see the dregs removed from pppd, but there was a patch release on the netatalk package just last month, so what do I know?
I think netatalk 3.0 dropped all appletalk protocol stuff a long time ago and only supports AFP over IP.
That's right. The older netatalk 2.x branch has a few different branches on GitHub that are being maintained to allow AppleTalk connectivity with older Macs through Ethernet (or LocalTalk-to-Ethernet bridges), so people are still actively using AppleTalk with Ethernet.
BTW, I messed up my quoting in my previous message so it was hard to see part of my reply -- but I agree that everything in drivers/net/appletalk could go away except for the Kconfig entry for CONFIG_ATALK. And dev.c and the other LocalTalk-specific bits in net/appletalk could go away.