* [2.6 patch] improce USB Gadget Kconfig @ 2003-11-23 17:23 Adrian Bunk 2003-12-30 3:54 ` David Brownell 0 siblings, 1 reply; 7+ messages in thread From: Adrian Bunk @ 2003-11-23 17:23 UTC (permalink / raw) To: David Brownell; +Cc: linux-kernel, greg, linux-usb-devel The patch below contains small changes to the USB Gadget Kconfig. The main change is that multiple modular peripheral controllers are no longer allowed (currently only one is there, but this may change). cu Adrian --- linux-2.6.0-test9-mm4/drivers/usb/gadget/Kconfig.old 2003-11-23 17:17:51.000000000 +0100 +++ linux-2.6.0-test9-mm4/drivers/usb/gadget/Kconfig 2003-11-23 17:52:24.000000000 +0100 @@ -3,10 +3,9 @@ # (a) a peripheral controller, and # (b) the gadget driver using it. # -# for 2.5 kbuild, drivers/usb/gadget/Kconfig # source this at the end of drivers/usb/Kconfig # -menuconfig USB_GADGET +config USB_GADGET tristate "Support for USB Gadgets" depends on EXPERIMENTAL help @@ -32,16 +31,24 @@ If in doubt, say "N" and don't enable these drivers; most people don't have this kind of hardware (except maybe inside Linux PDAs). +menu "USB Gadget options" + depends on USB_GADGET!=n + # # USB Peripheral Controller Support # + +# +# only one gadget controller driver, linked statically or as a module +# (depending on whether USB_GADGET is y or m) +# + choice prompt "USB Peripheral Controller Support" - depends on USB_GADGET -config USB_NET2280 - tristate "NetChip 2280 USB Peripheral Controller" - depends on PCI && USB_GADGET +config USB_TMP_NET2280 + bool "NetChip 2280 USB Peripheral Controller" + depends on PCI help NetChip 2280 is a PCI based USB peripheral controller which supports both full and high speed USB 2.0 data transfers. @@ -56,19 +63,28 @@ endchoice +config USB_NET2280 + tristate + default USB_TMP_NET2280 && USB_GADGET + + # # USB Gadget Drivers # + +# +# at most one gadget driver statically linked, +# OR any number of gadget drivers, linked as modules +# + choice prompt "USB Gadget Drivers" depends on USB_GADGET default USB_ETH -# FIXME want a cleaner dependency/config approach for drivers. - config USB_ZERO tristate "Gadget Zero (DEVELOPMENT)" - depends on USB_GADGET && (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100) + depends on USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100 help Gadget Zero is a two-configuration device. It either sinks and sources bulk data; or it loops back a configurable number of @@ -110,7 +126,7 @@ config USB_ETH tristate "Ethernet Gadget" - depends on USB_GADGET && NET && (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100) + depends on NET && (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100) help This driver implements Ethernet style communication, in either of two ways: @@ -155,7 +171,7 @@ config USB_GADGETFS tristate "Gadget Filesystem (EXPERIMENTAL)" - depends on USB_GADGET && (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX) && EXPERIMENTAL + depends on (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX) && EXPERIMENTAL help This driver provides a filesystem based API that lets user mode programs implement a single-configuration USB device, including @@ -179,4 +195,4 @@ endchoice -# endmenuconfig +endmenu ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [2.6 patch] improce USB Gadget Kconfig 2003-11-23 17:23 [2.6 patch] improce USB Gadget Kconfig Adrian Bunk @ 2003-12-30 3:54 ` David Brownell 2004-01-10 3:04 ` Adrian Bunk 2004-01-15 21:55 ` Roman Zippel 0 siblings, 2 replies; 7+ messages in thread From: David Brownell @ 2003-12-30 3:54 UTC (permalink / raw) To: Adrian Bunk; +Cc: linux-kernel, greg, linux-usb-devel, Roman Zippel [-- Attachment #1: Type: text/plain, Size: 889 bytes --] Adrian Bunk wrote: > The patch below contains small changes to the USB Gadget Kconfig. > > The main change is that multiple modular peripheral controllers are no > longer allowed (currently only one is there, but this may change). How about using this approach instead? It simplifies the kconfig for the gadget drivers by providing a boolean "which hardware" symbol, so gadget drivers don't need to make their own. The symbol that's synthetic is the one needed only by the Makefile. It's a larger change, but likely a better one. Greg, does it behave for you? Roman, this seems to trigger some kind of xconfig/menuconfig bug, since I can go down the list of hardware options (net2280, goku, dummy -- three, not the single one Adrian was working with) and each deselects the previous selection ... but then it's impossible to turn off the dummy, and select real hardware. - Dave [-- Attachment #2: Kconfig --] [-- Type: text/plain, Size: 9645 bytes --] # # USB Gadget support on a system involves # (a) a peripheral controller, and # (b) the gadget driver using it. # menu "USB Gadget Support" config USB_GADGET tristate "Support for USB Gadgets" depends on EXPERIMENTAL help USB is a master/slave protocol, organized with one master host (such as a PC) controlling up to 127 peripheral devices. The USB hardware is asymmetric, which makes it easier to set up: you can't connect two "to-the-host" connectors to each other. Linux can run in the host, or in the peripheral. In both cases you need a low level bus controller driver, and some software talking to it. Peripheral controllers are often discrete silicon, or are integrated with the CPU in a microcontroller. The more familiar host side controllers have names like like "EHCI", "OHCI", or "UHCI", and are usually integrated into southbridges on PC motherboards. Enable this configuration option if you want to run Linux inside a USB peripheral device. Configure one hardware driver for your peripheral/device side bus controller, and a "gadget driver" for your peripheral protocol. (If you use modular gadget drivers, you may configure more than one.) If in doubt, say "N" and don't enable these drivers; most people don't have this kind of hardware (except maybe inside Linux PDAs). # # USB Peripheral Controller Support # choice prompt "USB Peripheral Controller Support" depends on USB_GADGET != n optional help A USB device uses a controller to talk to its host. Systems should have only one such upstream link. config USB_GADGET_NET2280 boolean "NetChip 2280 USB Peripheral Controller" depends on PCI help NetChip 2280 is a PCI based USB peripheral controller which supports both full and high speed USB 2.0 data transfers. It has six configurable endpoints, as well as endpoint zero (for control transfers) and several endpoints with dedicated functions. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "net2280" and force all gadget drivers to also be dynamically linked. config USB_GADGET_PXA2XX boolean "PXA 2xx USB Device Controller" depends on ARCH_PXA help Intel's PXA 2xx series XScale ARM-5TE processors include an integrated full speed USB 1.1 device controller. It has fifteen fixed-function endpoints, as well as endpoint zero (for control transfers). Say "y" to link the driver statically, or "m" to build a dynamically linked module called "pxa2xx_udc" and force all gadget drivers to also be dynamically linked. config USB_GADGET_GOKU boolean "Toshiba TC86C001 (Goku-S) USB Device Controller" depends on PCI help The Toshiba TC86C001 is a PCI device which includes controllers for full speed USB devices, IDE, I2C, SIO, plus a USB host (OHCI). The device controller has three configurable (bulk or interrupt) endpoints, plus endpoint zero (for control transfers). Say "y" to link the driver statically, or "m" to build a dynamically linked module called "goku_udc" and to force all gadget drivers to also be dynamically linked. # this could be built elsewhere (doesn't yet exist) config USB_GADGET_SA1100 boolean "SA 1100 USB Device Controller" depends on ARCH_SA1100 help Intel's SA-1100 is an ARM-4 processor with an integrated full speed USB 1.1 device controller. It has two fixed-function endpoints, as well as endpoint zero (for control transfers). config USB_GADGET_DUMMY_HCD boolean "Dummy HCD support (DEVELOPMENT)" depends on USB help This host controller driver emulates USB, looping all data transfer requests back to a USB "gadget driver" in the same host. The host side is the master; the gadget side is the slave. Gadget drivers can be high, full, or low speed; and they have access to endpoints like those from NET2280, PXA2xx, or SA1100 hardware. This may help in some stages of creating a driver to embed in a Linux device, since it lets you debug several parts of the gadget driver without its hardware or drivers being involved. Since such a gadget side driver needs to interoperate with a host side Linux-USB device driver, this may help to debug both sides of a USB protocol stack. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "dummy_hcd" and force all gadget drivers to also be dynamically linked. endchoice # Gadget drivers should use the CONFIG_USB_GADGET_XXX symbol # for hardware-specific logic (like driver configuration). # These sibling CONFIG_USB_XXX symbols are for the Makefile. config USB_NET2280 tristate default n if USB_GADGET_NET2280 = n default m if USB_GADGET = m default y config USB_PXA2XX tristate default n if USB_GADGET_PXA2XX = n default m if USB_GADGET = m default y # if there's only one gadget driver, using only two bulk endpoints, # the other endpoints don't need to waste that extra memory config USB_PXA2XX_SMALL depends on USB_PXA2XX bool default y if (USB_ZERO = y) && (USB_PXA2XX = y) default y if (USB_ETH = y) && (USB_PXA2XX = y) default y if (USB_G_SERIAL = y) && (USB_PXA2XX = y) default n config USB_GOKU tristate default n if USB_GADGET_GOKU = n default m if USB_GADGET = m default y config USB_SA1100 tristate default n if USB_GADGET_SA1100 = n default m if USB_GADGET = m default y config USB_DUMMY_HCD tristate default n if USB_GADGET_DUMMY_HCD = n default m if USB_GADGET = m default y # # USB Gadget Drivers # choice tristate "USB Gadget Drivers" depends on USB_GADGET default USB_ETH config USB_ZERO tristate "Gadget Zero (DEVELOPMENT)" depends USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_GOKU || USB_SA1100 help Gadget Zero is a two-configuration device. It either sinks and sources bulk data; or it loops back a configurable number of transfers. It also implements control requests, for "chapter 9" conformance. The driver needs only two bulk-capable endpoints, so it can work on top of most device-side usb controllers. It's useful for testing, and is also a working example showing how USB "gadget drivers" can be written. Make this be the first driver you try using on top of any new USB peripheral controller driver. Then you can use host-side test software, like the "usbtest" driver, to put your hardware and its driver through a basic set of functional tests. Gadget Zero also works with the host-side "usb-skeleton" driver, and with many kinds of host-side test software. You may need to tweak product and vendor IDs before host software knows about this device, and arrange to select an appropriate configuration. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_zero". config USB_ETH tristate "Ethernet Gadget" depends on NET && (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_GOKU || USB_SA1100) help This driver implements Ethernet style communication, in either of two ways: - The "Communication Device Class" (CDC) Ethernet Control Model. That protocol is often avoided with pure Ethernet adapters, in favor of simpler vendor-specific hardware, but is widely supported by firmware for smart network devices. - On hardware can't implement that protocol, a simpler approach is used, placing fewer demands on USB. Within the USB device, this gadget driver exposes a network device "usbX", where X depends on what other networking devices you have. Treat it like a two-node Ethernet link: host, and gadget. The Linux-USB host-side "usbnet" driver interoperates with this driver, so that deep I/O queues can be supported. On 2.4 kernels, use "CDCEther" instead, if you're using the CDC option. That CDC mode should also interoperate with standard CDC Ethernet class drivers on other host operating systems. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_ether". config USB_GADGETFS tristate "Gadget Filesystem (EXPERIMENTAL)" depends on (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100 || USB_GOKU) && EXPERIMENTAL help This driver provides a filesystem based API that lets user mode programs implement a single-configuration USB device, including endpoint I/O and control requests that don't relate to enumeration. All endpoints, transfer speeds, and transfer types supported by the hardware are available, through read() and write() calls. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "gadgetfs". config USB_FILE_STORAGE tristate "File-backed Storage Gadget (DEVELOPMENT)" depends on USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_GOKU # we don't support the SA1100 because of its limitations help The File-backed Storage Gadget acts as a USB Mass Storage disk drive. As its storage repository it can use a regular file or a block device (in much the same way as the "loop" device driver), specified as a module parameter. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_file_storage". config USB_G_SERIAL tristate "Serial Gadget" depends on USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_GOKU || USB_SA1100 help The Serial Gadget talks to the Linux-USB generic serial driver. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_serial". endchoice endmenu ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [2.6 patch] improce USB Gadget Kconfig 2003-12-30 3:54 ` David Brownell @ 2004-01-10 3:04 ` Adrian Bunk 2004-01-15 21:55 ` Roman Zippel 1 sibling, 0 replies; 7+ messages in thread From: Adrian Bunk @ 2004-01-10 3:04 UTC (permalink / raw) To: David Brownell; +Cc: linux-kernel, greg, linux-usb-devel, Roman Zippel Hi David, first of all sorry for my late answer. I'm currently wading through a bunch of mails in my INBOX that should have been answered some time ago. :-( On Mon, Dec 29, 2003 at 07:54:29PM -0800, David Brownell wrote: > Adrian Bunk wrote: > >The patch below contains small changes to the USB Gadget Kconfig. > > > >The main change is that multiple modular peripheral controllers are no > >longer allowed (currently only one is there, but this may change). > > How about using this approach instead? It simplifies the kconfig > for the gadget drivers by providing a boolean "which hardware" > symbol, so gadget drivers don't need to make their own. The symbol > that's synthetic is the one needed only by the Makefile. >... Your suggestion looks fine. > - Dave >... cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [2.6 patch] improce USB Gadget Kconfig 2003-12-30 3:54 ` David Brownell 2004-01-10 3:04 ` Adrian Bunk @ 2004-01-15 21:55 ` Roman Zippel 2004-01-16 2:18 ` David Brownell 1 sibling, 1 reply; 7+ messages in thread From: Roman Zippel @ 2004-01-15 21:55 UTC (permalink / raw) To: David Brownell; +Cc: Adrian Bunk, linux-kernel, greg, linux-usb-devel Hi, On Mon, 29 Dec 2003, David Brownell wrote: > How about using this approach instead? It simplifies the kconfig > for the gadget drivers by providing a boolean "which hardware" > symbol, so gadget drivers don't need to make their own. The symbol > that's synthetic is the one needed only by the Makefile. There are some strange things in there. choice values can also be tristate symbols, so you wouldn't need the separate defines, unless you really always want to compile only a single controller (even as module). The "default m if USB_GADGET = m" looks weird, if I understand them correctly this should just be "depends on USB_GADGET", e.g. config USB_NET2280 tristate depends on USB_GADGET default USB_GADGET_NET2280 this would also fix the menu structure and the drivers menu would appear below the gadget option. I'm also not sure about USB_PXA2XX_SMALL, as it also can be written as: config USB_PXA2XX_SMALL depends on USB_PXA2XX = y default USB_ZERO = y || USB_ETH = y || USB_G_SERIAL is this really intended? The dependency "USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100 || USB_GOKU" can be basically reduced to "USB_GADGET". > Roman, this seems to trigger some kind of xconfig/menuconfig bug, > since I can go down the list of hardware options (net2280, goku, > dummy -- three, not the single one Adrian was working with) and > each deselects the previous selection ... but then it's impossible > to turn off the dummy, and select real hardware. I can't reproduce this, it works fine here. bye, Roman ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [2.6 patch] improce USB Gadget Kconfig 2004-01-15 21:55 ` Roman Zippel @ 2004-01-16 2:18 ` David Brownell 2004-01-16 21:41 ` Roman Zippel 0 siblings, 1 reply; 7+ messages in thread From: David Brownell @ 2004-01-16 2:18 UTC (permalink / raw) To: Roman Zippel; +Cc: Adrian Bunk, linux-kernel, greg, linux-usb-devel [-- Attachment #1: Type: text/plain, Size: 2659 bytes --] Roman Zippel wrote: > Hi, > > On Mon, 29 Dec 2003, David Brownell wrote: > >> How about using this approach instead? It simplifies the kconfig >> for the gadget drivers .. > > There are some strange things in there. Some of it was to kick the menu layout heuristic into something closer to sanity. There was also baggage from less-successful attempts to make the configuration behave right. > choice values can also be tristate symbols, so you wouldn't need the > separate defines, unless you really always want to compile only a single > controller (even as module). That's it precisely. USB devices have only one (upstream) link; they're not like hosts. And its link to the controller isn't re-wired on the fly any more than, say, the MMU. Kconfig just needed some persuasion before it'd dance that way. > The "default m if USB_GADGET = m" looks weird, if I understand them > correctly this should just be "depends on USB_GADGET", e.g. > > config USB_NET2280 > tristate > depends on USB_GADGET > default USB_GADGET_NET2280 > > this would also fix the menu structure and the drivers menu would appear > below the gadget option. More like config USB_GOKU tristate depends on USB_GADGET_GOKU default USB_GADGET And similar for net2280, pxa2xx, and so on. Either that, or moving it up higher in the text file, seems to have been the black magic that made the menu layout code behave. > I'm also not sure about USB_PXA2XX_SMALL, as it also can be written as: > > config USB_PXA2XX_SMALL > depends on USB_PXA2XX = y > default USB_ZERO = y || USB_ETH = y || USB_G_SERIAL > > is this really intended? I'm not sure what you're asking. I wrote it with one line per driver that's less error-prone in case updates get merged. The latest version is more terse, but there are lots of ways to write that kind of logic. > The dependency "USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100 > || USB_GOKU" can be basically reduced to "USB_GADGET". > >> Roman, this seems to trigger some kind of xconfig/menuconfig bug, >> since I can go down the list of hardware options (net2280, goku, >> dummy -- three, not the single one Adrian was working with) and >> each deselects the previous selection ... but then it's impossible >> to turn off the dummy, and select real hardware. > > > I can't reproduce this, it works fine here. Reproduced it again here today, with a reasonably current 2.6.1 tree on top of RH9 (plus some updated RPMs from RH). It's there in gconfig too. The workaround is "vi .config" and delete the sticky DUMMY_HCD entry, then re-configure. - Dave > bye, Roman > [-- Attachment #2: Kconfig --] [-- Type: text/plain, Size: 9589 bytes --] # # USB Gadget support on a system involves # (a) a peripheral controller, and # (b) the gadget driver using it. # menu "USB Gadget Support" config USB_GADGET tristate "Support for USB Gadgets" help USB is a master/slave protocol, organized with one master host (such as a PC) controlling up to 127 peripheral devices. The USB hardware is asymmetric, which makes it easier to set up: you can't connect two "to-the-host" connectors to each other. Linux can run in the host, or in the peripheral. In both cases you need a low level bus controller driver, and some software talking to it. Peripheral controllers are often discrete silicon, or are integrated with the CPU in a microcontroller. The more familiar host side controllers have names like like "EHCI", "OHCI", or "UHCI", and are usually integrated into southbridges on PC motherboards. Enable this configuration option if you want to run Linux inside a USB peripheral device. Configure one hardware driver for your peripheral/device side bus controller, and a "gadget driver" for your peripheral protocol. (If you use modular gadget drivers, you may configure more than one.) If in doubt, say "N" and don't enable these drivers; most people don't have this kind of hardware (except maybe inside Linux PDAs). # # USB Peripheral Controller Support # choice prompt "USB Peripheral Controller" depends on USB_GADGET help A USB device uses a controller to talk to its host. Systems should have only one such upstream link. config USB_GADGET_NET2280 boolean "NetChip 2280" depends on PCI help NetChip 2280 is a PCI based USB peripheral controller which supports both full and high speed USB 2.0 data transfers. It has six configurable endpoints, as well as endpoint zero (for control transfers) and several endpoints with dedicated functions. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "net2280" and force all gadget drivers to also be dynamically linked. config USB_NET2280 tristate depends on USB_GADGET_NET2280 default USB_GADGET config USB_GADGET_PXA2XX boolean "PXA 2xx or IXP 42x" depends on ARCH_PXA || ARCH_IXP425 help Intel's PXA 2xx series XScale ARM-5TE processors include an integrated full speed USB 1.1 device controller. The controller in the IXP 4xx series is register-compatible. It has fifteen fixed-function endpoints, as well as endpoint zero (for control transfers). Say "y" to link the driver statically, or "m" to build a dynamically linked module called "pxa2xx_udc" and force all gadget drivers to also be dynamically linked. config USB_PXA2XX tristate depends on USB_GADGET_PXA2XX default USB_GADGET # if there's only one gadget driver, using only two bulk endpoints, # don't waste memory for the other endpoints config USB_PXA2XX_SMALL depends on USB_GADGET_PXA2XX bool default y if (USB_ZERO = y) default y if (USB_ETH = y) default y if (USB_G_SERIAL = y) config USB_GADGET_GOKU boolean "Toshiba TC86C001 'Goku-S'" depends on PCI help The Toshiba TC86C001 is a PCI device which includes controllers for full speed USB devices, IDE, I2C, SIO, plus a USB host (OHCI). The device controller has three configurable (bulk or interrupt) endpoints, plus endpoint zero (for control transfers). Say "y" to link the driver statically, or "m" to build a dynamically linked module called "goku_udc" and to force all gadget drivers to also be dynamically linked. config USB_GOKU tristate depends on USB_GADGET_GOKU default USB_GADGET # this could be built elsewhere (doesn't yet exist) config USB_GADGET_SA1100 boolean "SA 1100" depends on ARCH_SA1100 help Intel's SA-1100 is an ARM-4 processor with an integrated full speed USB 1.1 device controller. It has two fixed-function endpoints, as well as endpoint zero (for control transfers). config USB_SA1100 tristate depends on USB_GADGET_SA1100 default USB_GADGET config USB_GADGET_DUMMY_HCD boolean "Dummy HCD (DEVELOPMENT)" depends on USB help This host controller driver emulates USB, looping all data transfer requests back to a USB "gadget driver" in the same host. The host side is the master; the gadget side is the slave. Gadget drivers can be high, full, or low speed; and they have access to endpoints like those from NET2280, PXA2xx, or SA1100 hardware. This may help in some stages of creating a driver to embed in a Linux device, since it lets you debug several parts of the gadget driver without its hardware or drivers being involved. Since such a gadget side driver needs to interoperate with a host side Linux-USB device driver, this may help to debug both sides of a USB protocol stack. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "dummy_hcd" and force all gadget drivers to also be dynamically linked. config USB_DUMMY_HCD tristate depends on USB_GADGET_DUMMY_HCD default USB_GADGET endchoice # # USB Gadget Drivers # choice tristate "USB Gadget Drivers" depends on USB_GADGET default USB_ETH # FIXME Gadget drivers should now just #ifdef CONFIG_USB_GADGET_XXX; # remove these other hardware flags config USB_ZERO tristate "Gadget Zero (DEVELOPMENT)" depends on (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_GOKU || USB_SA1100) help Gadget Zero is a two-configuration device. It either sinks and sources bulk data; or it loops back a configurable number of transfers. It also implements control requests, for "chapter 9" conformance. The driver needs only two bulk-capable endpoints, so it can work on top of most device-side usb controllers. It's useful for testing, and is also a working example showing how USB "gadget drivers" can be written. Make this be the first driver you try using on top of any new USB peripheral controller driver. Then you can use host-side test software, like the "usbtest" driver, to put your hardware and its driver through a basic set of functional tests. Gadget Zero also works with the host-side "usb-skeleton" driver, and with many kinds of host-side test software. You may need to tweak product and vendor IDs before host software knows about this device, and arrange to select an appropriate configuration. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_zero". config USB_ETH tristate "Ethernet Gadget" depends on NET && (USB_NET2280 || USB_PXA2XX || USB_GOKU || USB_SA1100) help This driver implements Ethernet style communication, in either of two ways: - The "Communication Device Class" (CDC) Ethernet Control Model. That protocol is often avoided with pure Ethernet adapters, in favor of simpler vendor-specific hardware, but is widely supported by firmware for smart network devices. - On hardware can't implement that protocol, a simpler approach is used, placing fewer demands on USB. Within the USB device, this gadget driver exposes a network device "usbX", where X depends on what other networking devices you have. Treat it like a two-node Ethernet link: host, and gadget. The Linux-USB host-side "usbnet" driver interoperates with this driver, so that deep I/O queues can be supported. On 2.4 kernels, use "CDCEther" instead, if you're using the CDC option. That CDC mode should also interoperate with standard CDC Ethernet class drivers on other host operating systems. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_ether". config USB_GADGETFS tristate "Gadget Filesystem (EXPERIMENTAL)" depends on (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100 || USB_GOKU) && EXPERIMENTAL help This driver provides a filesystem based API that lets user mode programs implement a single-configuration USB device, including endpoint I/O and control requests that don't relate to enumeration. All endpoints, transfer speeds, and transfer types supported by the hardware are available, through read() and write() calls. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "gadgetfs". config USB_FILE_STORAGE tristate "File-backed Storage Gadget (DEVELOPMENT)" depends on (USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_GOKU) # we don't support the SA1100 because of its limitations help The File-backed Storage Gadget acts as a USB Mass Storage disk drive. As its storage repository it can use a regular file or a block device (in much the same way as the "loop" device driver), specified as a module parameter. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_file_storage". config USB_FILE_STORAGE_TEST bool "File-backed Storage Gadget test version" depends on USB_FILE_STORAGE default n help Say "y" to generate the larger testing version of the File-backed Storage Gadget, useful for probing the behavior of USB Mass Storage hosts. Not needed for normal operation. config USB_G_SERIAL tristate "Serial Gadget" depends on (USB_NET2280 || USB_PXA2XX || USB_GOKU || USB_SA1100) help The Serial Gadget talks to the Linux-USB generic serial driver. Say "y" to link the driver statically, or "m" to build a dynamically linked module called "g_serial". endchoice endmenu ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [2.6 patch] improce USB Gadget Kconfig 2004-01-16 2:18 ` David Brownell @ 2004-01-16 21:41 ` Roman Zippel 2004-01-16 23:15 ` David Brownell 0 siblings, 1 reply; 7+ messages in thread From: Roman Zippel @ 2004-01-16 21:41 UTC (permalink / raw) To: David Brownell; +Cc: Adrian Bunk, linux-kernel, greg, linux-usb-devel Hi, On Thu, 15 Jan 2004, David Brownell wrote: > > choice values can also be tristate symbols, so you wouldn't need the > > separate defines, unless you really always want to compile only a single > > controller (even as module). > > That's it precisely. USB devices have only one (upstream) link; > they're not like hosts. And its link to the controller isn't > re-wired on the fly any more than, say, the MMU. Kconfig just > needed some persuasion before it'd dance that way. It's still weird. Where is the problem to compile all controllers as module? At runtime you still have the possibility that only one of them can be loaded and the next one would fail to load. > And similar for net2280, pxa2xx, and so on. Either that, or moving it > up higher in the text file, seems to have been the black magic that > made the menu layout code behave. To fix the menu layout it's probably the easiest to put most of it within a "if USB_GADGET" ... "endif". > > I'm also not sure about USB_PXA2XX_SMALL, as it also can be written as: > > > > config USB_PXA2XX_SMALL > > depends on USB_PXA2XX = y > > default USB_ZERO = y || USB_ETH = y || USB_G_SERIAL > > > > is this really intended? > > I'm not sure what you're asking. I wrote it with one line per > driver that's less error-prone in case updates get merged. The > latest version is more terse, but there are lots of ways to > write that kind of logic. The comment confuses me, I don't see how it tests that it's "only one" of something. > Reproduced it again here today, with a reasonably current 2.6.1 > tree on top of RH9 (plus some updated RPMs from RH). It's there > in gconfig too. The workaround is "vi .config" and delete the > sticky DUMMY_HCD entry, then re-configure. It really works fine here, are you sure you don't have any additional changes under scripts/kconfig? Did you try this on a different machine? bye, Roman ^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [2.6 patch] improce USB Gadget Kconfig 2004-01-16 21:41 ` Roman Zippel @ 2004-01-16 23:15 ` David Brownell 0 siblings, 0 replies; 7+ messages in thread From: David Brownell @ 2004-01-16 23:15 UTC (permalink / raw) To: Roman Zippel; +Cc: Adrian Bunk, linux-kernel, greg, linux-usb-devel Roman Zippel wrote: > Hi, > > On Thu, 15 Jan 2004, David Brownell wrote: > >> > choice values can also be tristate symbols, so you wouldn't need the >> > separate defines, unless you really always want to compile only a >> > single controller (even as module). >> >> That's it precisely. USB devices have only one (upstream) link; >> they're not like hosts. And its link to the controller isn't >> re-wired on the fly any more than, say, the MMU. Kconfig just >> needed some persuasion before it'd dance that way. > > > It's still weird. Where is the problem to compile all controllers as > module? At runtime you still have the possibility that only one of them > can be loaded and the next one would fail to load. The problem is something I accidently neglected to respond to in your previous note. Those controllers aren't drop-in substitutes for one another, and you seemed to be assuming they were. >> > The dependency "USB_DUMMY_HCD || USB_NET2280 || USB_PXA2XX || USB_SA1100 >> > || USB_GOKU" can be basically reduced to "USB_GADGET". That's where I accidentally trimmed out a response. It's wrong; those controllers are not equivalent in general. Gadget drivers that work with one controller will not automagically work with another. Two quick examples should illustrate the point: (i) A "video" gadget driver would requires isochronous transfer support. But only two of those controllers support that (net2280, pxa2xx_udc). So a video (or audio) class gadget driver would depend on configuring an appropriate controller ... and even then, it'd probably compile in more code if it's talking to a controller that supports high bandwidth transfer modes (up to 24 MByte/sec) than one maxing ouat a 1 Mbyte/sec. (ii) Essentially every controller has endpoints that differ from the others in capabilities like addressing, direction, supported packet sizes transfer modes, or throughput. All of those are things that MUST be reported to the USB host in device descriptors. The differences are currently all handled by conditional compilation. Until a gadget driver has that support for a particular controller, it shouldn't be possible to configure that combination. There's some work afoot to do more autoconfiguration, but it won't be able to handle all the issues for all drivers. I have hopes that at least some simple configurations -- two endpoints for bulk data (IN/OUT) and one for status (interrupt IN), for example -- will be able to autoconfigure at some point. >> Reproduced it again here today, with a reasonably current 2.6.1 >> tree on top of RH9 (plus some updated RPMs from RH). It's there >> in gconfig too. The workaround is "vi .config" and delete the >> sticky DUMMY_HCD entry, then re-configure. > > > It really works fine here, are you sure you don't have any additional > changes under scripts/kconfig? Did you try this on a different machine? Absolutely certain. Yes, tried on multiple machines. Did you try it on RH9? Since it's working for you, I wonder if that's the source of the difference. Same bug in xconfig, gconfig, and menuconfig; doesn't make a lot of sense to me but then I've not looked at that code in any case. - Dave ^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2004-01-16 23:15 UTC | newest] Thread overview: 7+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2003-11-23 17:23 [2.6 patch] improce USB Gadget Kconfig Adrian Bunk 2003-12-30 3:54 ` David Brownell 2004-01-10 3:04 ` Adrian Bunk 2004-01-15 21:55 ` Roman Zippel 2004-01-16 2:18 ` David Brownell 2004-01-16 21:41 ` Roman Zippel 2004-01-16 23:15 ` David Brownell
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox