linux_old1/drivers/net/wireless/wavelan.p.h

705 lines
26 KiB
C
Raw Normal View History

/*
* WaveLAN ISA driver
*
* Jean II - HPLB '96
*
* Reorganisation and extension of the driver.
*
* This file contains all definitions and declarations necessary for the
* WaveLAN ISA driver. This file is a private header, so it should
* be included only in wavelan.c!
*/
#ifndef WAVELAN_P_H
#define WAVELAN_P_H
/************************** DOCUMENTATION ***************************/
/*
* This driver provides a Linux interface to the WaveLAN ISA hardware.
* The WaveLAN is a product of Lucent (http://www.wavelan.com/).
* This division was formerly part of NCR and then AT&T.
* WaveLANs are also distributed by DEC (RoamAbout DS) and Digital Ocean.
*
* To learn how to use this driver, read the NET3 HOWTO.
* If you want to exploit the many other functionalities, read the comments
* in the code.
*
* This driver is the result of the effort of many people (see below).
*/
/* ------------------------ SPECIFIC NOTES ------------------------ */
/*
* Web page
* --------
* I try to maintain a web page with the Wireless LAN Howto at :
* http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux/Wavelan.html
*
* SMP
* ---
* We now are SMP compliant (I eventually fixed the remaining bugs).
* The driver has been tested on a dual P6-150 and survived my usual
* set of torture tests.
* Anyway, I spent enough time chasing interrupt re-entrancy during
* errors or reconfigure, and I designed the locked/unlocked sections
* of the driver with great care, and with the recent addition of
* the spinlock (thanks to the new API), we should be quite close to
* the truth.
* The SMP/IRQ locking is quite coarse and conservative (i.e. not fast),
* but better safe than sorry (especially at 2 Mb/s ;-).
*
* I have also looked into disabling only our interrupt on the card
* (via HACR) instead of all interrupts in the processor (via cli),
* so that other driver are not impacted, and it look like it's
* possible, but it's very tricky to do right (full of races). As
* the gain would be mostly for SMP systems, it can wait...
*
* Debugging and options
* ---------------------
* You will find below a set of '#define" allowing a very fine control
* on the driver behaviour and the debug messages printed.
* The main options are :
* o SET_PSA_CRC, to have your card correctly recognised by
* an access point and the Point-to-Point diagnostic tool.
* o USE_PSA_CONFIG, to read configuration from the PSA (EEprom)
* (otherwise we always start afresh with some defaults)
*
* wavelan.o is too darned big
* ---------------------------
* That's true! There is a very simple way to reduce the driver
* object by 33%! Comment out the following line:
* #include <linux/wireless.h>
* Other compile options can also reduce the size of it...
*
* MAC address and hardware detection:
* -----------------------------------
* The detection code for the WaveLAN checks that the first three
* octets of the MAC address fit the company code. This type of
* detection works well for AT&T cards (because the AT&T code is
* hardcoded in wavelan.h), but of course will fail for other
* manufacturers.
*
* If you are sure that your card is derived from the WaveLAN,
* here is the way to configure it:
* 1) Get your MAC address
* a) With your card utilities (wfreqsel, instconf, etc.)
* b) With the driver:
* o compile the kernel with DEBUG_CONFIG_INFO enabled
* o Boot and look the card messages
* 2) Set your MAC code (3 octets) in MAC_ADDRESSES[][3] (wavelan.h)
* 3) Compile and verify
* 4) Send me the MAC code. I will include it in the next version.
*
*/
/* --------------------- WIRELESS EXTENSIONS --------------------- */
/*
* This driver is the first to support "wireless extensions".
* This set of extensions provides a standard way to control the wireless
* characteristics of the hardware. Applications such as mobile IP may
* take advantage of it.
*
* It might be a good idea as well to fetch the wireless tools to
* configure the device and play a bit.
*/
/* ---------------------------- FILES ---------------------------- */
/*
* wavelan.c: actual code for the driver: C functions
*
* wavelan.p.h: private header: local types and variables for driver
*
* wavelan.h: description of the hardware interface and structs
*
* i82586.h: description of the Ethernet controller
*/
/* --------------------------- HISTORY --------------------------- */
/*
* This is based on information in the drivers' headers. It may not be
* accurate, and I guarantee only my best effort.
*
* The history of the WaveLAN drivers is as complicated as the history of
* the WaveLAN itself (NCR -> AT&T -> Lucent).
*
* It all started with Anders Klemets <klemets@paul.rutgers.edu>
* writing a WaveLAN ISA driver for the Mach microkernel. Girish
* Welling <welling@paul.rutgers.edu> had also worked on it.
* Keith Moore modified this for the PCMCIA hardware.
*
* Robert Morris <rtm@das.harvard.edu> ported these two drivers to BSDI
* and added specific PCMCIA support (there is currently no equivalent
* of the PCMCIA package under BSD).
*
* Jim Binkley <jrb@cs.pdx.edu> ported both BSDI drivers to FreeBSD.
*
* Bruce Janson <bruce@cs.usyd.edu.au> ported the BSDI ISA driver to Linux.
*
* Anthony D. Joseph <adj@lcs.mit.edu> started to modify Bruce's driver
* (with help of the BSDI PCMCIA driver) for PCMCIA.
* Yunzhou Li <yunzhou@strat.iol.unh.edu> finished this work.
* Joe Finney <joe@comp.lancs.ac.uk> patched the driver to start
* 2.00 cards correctly (2.4 GHz with frequency selection).
* David Hinds <dahinds@users.sourceforge.net> integrated the whole in his
* PCMCIA package (and bug corrections).
*
* I (Jean Tourrilhes - jt@hplb.hpl.hp.com) then started to make some
* patches to the PCMCIA driver. Later, I added code in the ISA driver
* for Wireless Extensions and full support of frequency selection
* cards. Then, I did the same to the PCMCIA driver, and did some
* reorganisation. Finally, I came back to the ISA driver to
* upgrade it at the same level as the PCMCIA one and reorganise
* the code.
* Loeke Brederveld <lbrederv@wavelan.com> from Lucent has given me
* much needed information on the WaveLAN hardware.
*/
/* The original copyrights and literature mention others' names and
* credits. I don't know what their part in this development was.
*/
/* By the way, for the copyright and legal stuff:
* almost everybody wrote code under the GNU or BSD license (or similar),
* and want their original copyright to remain somewhere in the
* code (for myself, I go with the GPL).
* Nobody wants to take responsibility for anything, except the fame.
*/
/* --------------------------- CREDITS --------------------------- */
/*
* This software was developed as a component of the
* Linux operating system.
* It is based on other device drivers and information
* either written or supplied by:
* Ajay Bakre <bakre@paul.rutgers.edu>,
* Donald Becker <becker@cesdis.gsfc.nasa.gov>,
* Loeke Brederveld <Loeke.Brederveld@Utrecht.NCR.com>,
* Brent Elphick <belphick@uwaterloo.ca>,
* Anders Klemets <klemets@it.kth.se>,
* Vladimir V. Kolpakov <w@stier.koenig.ru>,
* Marc Meertens <Marc.Meertens@Utrecht.NCR.com>,
* Pauline Middelink <middelin@polyware.iaf.nl>,
* Robert Morris <rtm@das.harvard.edu>,
* Jean Tourrilhes <jt@hpl.hp.com>,
* Girish Welling <welling@paul.rutgers.edu>,
* Clark Woodworth <clark@hiway1.exit109.com>
* Yongguang Zhang <ygz@isl.hrl.hac.com>
*
* Thanks go also to:
* James Ashton <jaa101@syseng.anu.edu.au>,
* Alan Cox <alan@redhat.com>,
* Allan Creighton <allanc@cs.usyd.edu.au>,
* Matthew Geier <matthew@cs.usyd.edu.au>,
* Remo di Giovanni <remo@cs.usyd.edu.au>,
* Eckhard Grah <grah@wrcs1.urz.uni-wuppertal.de>,
* Vipul Gupta <vgupta@cs.binghamton.edu>,
* Mark Hagan <mhagan@wtcpost.daytonoh.NCR.COM>,
* Tim Nicholson <tim@cs.usyd.edu.au>,
* Ian Parkin <ian@cs.usyd.edu.au>,
* John Rosenberg <johnr@cs.usyd.edu.au>,
* George Rossi <george@phm.gov.au>,
* Arthur Scott <arthur@cs.usyd.edu.au>,
* Stanislav Sinyagin <stas@isf.ru>
* and Peter Storey for their assistance and advice.
*
* Additional Credits:
*
* My development has been done initially under Debian 1.1 (Linux 2.0.x)
* and now under Debian 2.2, initially with an HP Vectra XP/60, and now
* an HP Vectra XP/90.
*
*/
/* ------------------------- IMPROVEMENTS ------------------------- */
/*
* I proudly present:
*
* Changes made in first pre-release:
* ----------------------------------
* - reorganisation of the code, function name change
* - creation of private header (wavelan.p.h)
* - reorganised debug messages
* - more comments, history, etc.
* - mmc_init: configure the PSA if not done
* - mmc_init: correct default value of level threshold for PCMCIA
* - mmc_init: 2.00 detection better code for 2.00 initialization
* - better info at startup
* - IRQ setting (note: this setting is permanent)
* - watchdog: change strategy (and solve module removal problems)
* - add wireless extensions (ioctl and get_wireless_stats)
* get/set nwid/frequency on fly, info for /proc/net/wireless
* - more wireless extensions: SETSPY and GETSPY
* - make wireless extensions optional
* - private ioctl to set/get quality and level threshold, histogram
* - remove /proc/net/wavelan
* - suppress useless stuff from lp (net_local)
* - kernel 2.1 support (copy_to/from_user instead of memcpy_to/fromfs)
* - add message level (debug stuff in /var/adm/debug and errors not
* displayed at console and still in /var/adm/messages)
* - multi device support
* - start fixing the probe (init code)
* - more inlines
* - man page
* - many other minor details and cleanups
*
* Changes made in second pre-release:
* -----------------------------------
* - clean up init code (probe and module init)
* - better multiple device support (module)
* - name assignment (module)
*
* Changes made in third pre-release:
* ----------------------------------
* - be more conservative on timers
* - preliminary support for multicast (I still lack some details)
*
* Changes made in fourth pre-release:
* -----------------------------------
* - multicast (revisited and finished)
* - avoid reset in set_multicast_list (a really big hack)
* if somebody could apply this code for other i82586 based drivers
* - share onboard memory 75% RU and 25% CU (instead of 50/50)
*
* Changes made for release in 2.1.15:
* -----------------------------------
* - change the detection code for multi manufacturer code support
*
* Changes made for release in 2.1.17:
* -----------------------------------
* - update to wireless extensions changes
* - silly bug in card initial configuration (psa_conf_status)
*
* Changes made for release in 2.1.27 & 2.0.30:
* --------------------------------------------
* - small bug in debug code (probably not the last one...)
* - remove extern keyword for wavelan_probe()
* - level threshold is now a standard wireless extension (version 4 !)
* - modules parameters types (new module interface)
*
* Changes made for release in 2.1.36:
* -----------------------------------
* - byte count stats (courtesy of David Hinds)
* - remove dev_tint stuff (courtesy of David Hinds)
* - encryption setting from Brent Elphick (thanks a lot!)
* - 'ioaddr' to 'u_long' for the Alpha (thanks to Stanislav Sinyagin)
*
* Other changes (not by me) :
* -------------------------
* - Spelling and gramar "rectification".
*
* Changes made for release in 2.0.37 & 2.2.2 :
* ------------------------------------------
* - Correct status in /proc/net/wireless
* - Set PSA CRC to make PtP diagnostic tool happy (Bob Gray)
* - Module init code don't fail if we found at least one card in
* the address list (Karlis Peisenieks)
* - Missing parenthesis (Christopher Peterson)
* - Correct i82586 configuration parameters
* - Encryption initialisation bug (Robert McCormack)
* - New mac addresses detected in the probe
* - Increase watchdog for busy environments
*
* Changes made for release in 2.0.38 & 2.2.7 :
* ------------------------------------------
* - Correct the reception logic to better report errors and avoid
* sending bogus packet up the stack
* - Delay RU config to avoid corrupting first received packet
* - Change config completion code (to actually check something)
* - Avoid reading out of bound in skbuf to transmit
* - Rectify a lot of (useless) debugging code
* - Change the way to `#ifdef SET_PSA_CRC'
*
* Changes made for release in 2.2.11 & 2.3.13 :
* -------------------------------------------
* - Change e-mail and web page addresses
* - Watchdog timer is now correctly expressed in HZ, not in jiffies
* - Add channel number to the list of frequencies in range
* - Add the (short) list of bit-rates in range
* - Developp a new sensitivity... (sens.value & sens.fixed)
*
* Changes made for release in 2.2.14 & 2.3.23 :
* -------------------------------------------
* - Fix check for root permission (break instead of exit)
* - New nwid & encoding setting (Wireless Extension 9)
*
* Changes made for release in 2.3.49 :
* ----------------------------------
* - Indentation reformating (Alan)
* - Update to new network API (softnet - 2.3.43) :
* o replace dev->tbusy (Alan)
* o replace dev->tstart (Alan)
* o remove dev->interrupt (Alan)
* o add SMP locking via spinlock in splxx (me)
* o add spinlock in interrupt handler (me)
* o use kernel watchdog instead of ours (me)
* o increase watchdog timeout (kernel is more sensitive) (me)
* o verify that all the changes make sense and work (me)
* - Fixup a potential gotcha when reconfiguring and thighten a bit
* the interactions with Tx queue.
*
* Changes made for release in 2.4.0 :
* ---------------------------------
* - Fix spinlock stupid bugs that I left in. The driver is now SMP
* compliant and doesn't lockup at startup.
*
* Changes made for release in 2.5.2 :
* ---------------------------------
* - Use new driver API for Wireless Extensions :
* o got rid of wavelan_ioctl()
* o use a bunch of iw_handler instead
*
* Changes made for release in 2.5.35 :
* ----------------------------------
* - Set dev->trans_start to avoid filling the logs
* - Handle better spurious/bogus interrupt
* - Avoid deadlocks in mmc_out()/mmc_in()
*
* Wishes & dreams:
* ----------------
* - roaming (see Pcmcia driver)
*/
/***************************** INCLUDES *****************************/
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/types.h>
#include <linux/fcntl.h>
#include <linux/interrupt.h>
#include <linux/stat.h>
#include <linux/ptrace.h>
#include <linux/ioport.h>
#include <linux/in.h>
#include <linux/string.h>
#include <linux/delay.h>
#include <linux/bitops.h>
#include <asm/system.h>
#include <asm/io.h>
#include <asm/dma.h>
#include <asm/uaccess.h>
#include <linux/errno.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/slab.h>
#include <linux/timer.h>
#include <linux/init.h>
#include <linux/wireless.h> /* Wireless extensions */
#include <net/iw_handler.h> /* Wireless handlers */
/* WaveLAN declarations */
#include "i82586.h"
#include "wavelan.h"
/************************** DRIVER OPTIONS **************************/
/*
* `#define' or `#undef' the following constant to change the behaviour
* of the driver...
*/
#undef SET_PSA_CRC /* Calculate and set the CRC on PSA (slower) */
#define USE_PSA_CONFIG /* Use info from the PSA. */
#undef STRUCT_CHECK /* Verify padding of structures. */
#undef EEPROM_IS_PROTECTED /* doesn't seem to be necessary */
#define MULTICAST_AVOID /* Avoid extra multicast (I'm sceptical). */
#undef SET_MAC_ADDRESS /* Experimental */
/* Warning: this stuff will slow down the driver. */
#define WIRELESS_SPY /* Enable spying addresses. */
#undef HISTOGRAM /* Enable histogram of signal level. */
/****************************** DEBUG ******************************/
#undef DEBUG_MODULE_TRACE /* module insertion/removal */
#undef DEBUG_CALLBACK_TRACE /* calls made by Linux */
#undef DEBUG_INTERRUPT_TRACE /* calls to handler */
#undef DEBUG_INTERRUPT_INFO /* type of interrupt and so on */
#define DEBUG_INTERRUPT_ERROR /* problems */
#undef DEBUG_CONFIG_TRACE /* Trace the config functions. */
#undef DEBUG_CONFIG_INFO /* what's going on */
#define DEBUG_CONFIG_ERROR /* errors on configuration */
#undef DEBUG_TX_TRACE /* transmission calls */
#undef DEBUG_TX_INFO /* header of the transmitted packet */
#undef DEBUG_TX_FAIL /* Normal failure conditions */
#define DEBUG_TX_ERROR /* Unexpected conditions */
#undef DEBUG_RX_TRACE /* transmission calls */
#undef DEBUG_RX_INFO /* header of the received packet */
#undef DEBUG_RX_FAIL /* Normal failure conditions */
#define DEBUG_RX_ERROR /* Unexpected conditions */
#undef DEBUG_PACKET_DUMP /* Dump packet on the screen if defined to 32. */
#undef DEBUG_IOCTL_TRACE /* misc. call by Linux */
#undef DEBUG_IOCTL_INFO /* various debugging info */
#define DEBUG_IOCTL_ERROR /* what's going wrong */
#define DEBUG_BASIC_SHOW /* Show basic startup info. */
#undef DEBUG_VERSION_SHOW /* Print version info. */
#undef DEBUG_PSA_SHOW /* Dump PSA to screen. */
#undef DEBUG_MMC_SHOW /* Dump mmc to screen. */
#undef DEBUG_SHOW_UNUSED /* Show unused fields too. */
#undef DEBUG_I82586_SHOW /* Show i82586 status. */
#undef DEBUG_DEVICE_SHOW /* Show device parameters. */
/************************ CONSTANTS & MACROS ************************/
#ifdef DEBUG_VERSION_SHOW
static const char *version = "wavelan.c : v24 (SMP + wireless extensions) 11/12/01\n";
#endif
/* Watchdog temporisation */
#define WATCHDOG_JIFFIES (512*HZ/100)
/* ------------------------ PRIVATE IOCTL ------------------------ */
#define SIOCSIPQTHR SIOCIWFIRSTPRIV /* Set quality threshold */
#define SIOCGIPQTHR SIOCIWFIRSTPRIV + 1 /* Get quality threshold */
#define SIOCSIPHISTO SIOCIWFIRSTPRIV + 2 /* Set histogram ranges */
#define SIOCGIPHISTO SIOCIWFIRSTPRIV + 3 /* Get histogram values */
/****************************** TYPES ******************************/
/* Shortcuts */
typedef struct net_device_stats en_stats;
typedef struct iw_statistics iw_stats;
typedef struct iw_quality iw_qual;
typedef struct iw_freq iw_freq;
typedef struct net_local net_local;
typedef struct timer_list timer_list;
/* Basic types */
typedef u_char mac_addr[WAVELAN_ADDR_SIZE]; /* Hardware address */
/*
* Static specific data for the interface.
*
* For each network interface, Linux keeps data in two structures: "device"
* keeps the generic data (same format for everybody) and "net_local" keeps
* additional specific data.
* Note that some of this specific data is in fact generic (en_stats, for
* example).
*/
struct net_local
{
net_local * next; /* linked list of the devices */
struct net_device * dev; /* reverse link */
spinlock_t spinlock; /* Serialize access to the hardware (SMP) */
en_stats stats; /* Ethernet interface statistics */
int nresets; /* number of hardware resets */
u_char reconfig_82586; /* We need to reconfigure the controller. */
u_char promiscuous; /* promiscuous mode */
int mc_count; /* number of multicast addresses */
u_short hacr; /* current host interface state */
int tx_n_in_use;
u_short rx_head;
u_short rx_last;
u_short tx_first_free;
u_short tx_first_in_use;
iw_stats wstats; /* Wireless-specific statistics */
struct iw_spy_data spy_data;
struct iw_public_data wireless_data;
#ifdef HISTOGRAM
int his_number; /* number of intervals */
u_char his_range[16]; /* boundaries of interval ]n-1; n] */
u_long his_sum[16]; /* sum in interval */
#endif /* HISTOGRAM */
};
/**************************** PROTOTYPES ****************************/
/* ----------------------- MISC. SUBROUTINES ------------------------ */
static u_char
wv_irq_to_psa(int);
static int
wv_psa_to_irq(u_char);
/* ------------------- HOST ADAPTER SUBROUTINES ------------------- */
static inline u_short /* data */
hasr_read(u_long); /* Read the host interface: base address */
static inline void
hacr_write(u_long, /* Write to host interface: base address */
u_short), /* data */
hacr_write_slow(u_long,
u_short),
set_chan_attn(u_long, /* ioaddr */
u_short), /* hacr */
wv_hacr_reset(u_long), /* ioaddr */
wv_16_off(u_long, /* ioaddr */
u_short), /* hacr */
wv_16_on(u_long, /* ioaddr */
u_short), /* hacr */
wv_ints_off(struct net_device *),
wv_ints_on(struct net_device *);
/* ----------------- MODEM MANAGEMENT SUBROUTINES ----------------- */
static void
psa_read(u_long, /* Read the Parameter Storage Area. */
u_short, /* hacr */
int, /* offset in PSA */
u_char *, /* buffer to fill */
int), /* size to read */
psa_write(u_long, /* Write to the PSA. */
u_short, /* hacr */
int, /* offset in PSA */
u_char *, /* buffer in memory */
int); /* length of buffer */
static inline void
mmc_out(u_long, /* Write 1 byte to the Modem Manag Control. */
u_short,
u_char),
mmc_write(u_long, /* Write n bytes to the MMC. */
u_char,
u_char *,
int);
static inline u_char /* Read 1 byte from the MMC. */
mmc_in(u_long,
u_short);
static inline void
mmc_read(u_long, /* Read n bytes from the MMC. */
u_char,
u_char *,
int),
fee_wait(u_long, /* Wait for frequency EEPROM: base address */
int, /* base delay to wait for */
int); /* time to wait */
static void
fee_read(u_long, /* Read the frequency EEPROM: base address */
u_short, /* destination offset */
u_short *, /* data buffer */
int); /* number of registers */
/* ---------------------- I82586 SUBROUTINES ----------------------- */
static /*inline*/ void
obram_read(u_long, /* ioaddr */
u_short, /* o */
u_char *, /* b */
int); /* n */
static inline void
obram_write(u_long, /* ioaddr */
u_short, /* o */
u_char *, /* b */
int); /* n */
static void
wv_ack(struct net_device *);
static inline int
wv_synchronous_cmd(struct net_device *,
const char *),
wv_config_complete(struct net_device *,
u_long,
net_local *);
static int
wv_complete(struct net_device *,
u_long,
net_local *);
static inline void
wv_82586_reconfig(struct net_device *);
/* ------------------- DEBUG & INFO SUBROUTINES ------------------- */
#ifdef DEBUG_I82586_SHOW
static void
wv_scb_show(unsigned short);
#endif
static inline void
wv_init_info(struct net_device *); /* display startup info */
/* ------------------- IOCTL, STATS & RECONFIG ------------------- */
static en_stats *
wavelan_get_stats(struct net_device *); /* Give stats /proc/net/dev */
static iw_stats *
wavelan_get_wireless_stats(struct net_device *);
static void
wavelan_set_multicast_list(struct net_device *);
/* ----------------------- PACKET RECEPTION ----------------------- */
static inline void
wv_packet_read(struct net_device *, /* Read a packet from a frame. */
u_short,
int),
wv_receive(struct net_device *); /* Read all packets waiting. */
/* --------------------- PACKET TRANSMISSION --------------------- */
static inline int
wv_packet_write(struct net_device *, /* Write a packet to the Tx buffer. */
void *,
short);
static int
wavelan_packet_xmit(struct sk_buff *, /* Send a packet. */
struct net_device *);
/* -------------------- HARDWARE CONFIGURATION -------------------- */
static inline int
wv_mmc_init(struct net_device *), /* Initialize the modem. */
wv_ru_start(struct net_device *), /* Start the i82586 receiver unit. */
wv_cu_start(struct net_device *), /* Start the i82586 command unit. */
wv_82586_start(struct net_device *); /* Start the i82586. */
static void
wv_82586_config(struct net_device *); /* Configure the i82586. */
static inline void
wv_82586_stop(struct net_device *);
static int
wv_hw_reset(struct net_device *), /* Reset the WaveLAN hardware. */
wv_check_ioaddr(u_long, /* ioaddr */
u_char *); /* mac address (read) */
/* ---------------------- INTERRUPT HANDLING ---------------------- */
static irqreturn_t
wavelan_interrupt(int, /* interrupt handler */
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
void *);
static void
wavelan_watchdog(struct net_device *); /* transmission watchdog */
/* ------------------- CONFIGURATION CALLBACKS ------------------- */
static int
wavelan_open(struct net_device *), /* Open the device. */
wavelan_close(struct net_device *), /* Close the device. */
wavelan_config(struct net_device *, unsigned short);/* Configure one device. */
extern struct net_device *wavelan_probe(int unit); /* See Space.c. */
/**************************** VARIABLES ****************************/
/*
* This is the root of the linked list of WaveLAN drivers
* It is use to verify that we don't reuse the same base address
* for two different drivers and to clean up when removing the module.
*/
static net_local * wavelan_list = (net_local *) NULL;
/*
* This table is used to translate the PSA value to IRQ number
* and vice versa.
*/
static u_char irqvals[] =
{
0, 0, 0, 0x01,
0x02, 0x04, 0, 0x08,
0, 0, 0x10, 0x20,
0x40, 0, 0, 0x80,
};
/*
* Table of the available I/O addresses (base addresses) for WaveLAN
*/
static unsigned short iobase[] =
{
#if 0
/* Leave out 0x3C0 for now -- seems to clash with some video
* controllers.
* Leave out the others too -- we will always use 0x390 and leave
* 0x300 for the Ethernet device.
* Jean II: 0x3E0 is fine as well.
*/
0x300, 0x390, 0x3E0, 0x3C0
#endif /* 0 */
0x390, 0x3E0
};
#ifdef MODULE
/* Parameters set by insmod */
static int io[4];
static int irq[4];
static char *name[4];
module_param_array(io, int, NULL, 0);
module_param_array(irq, int, NULL, 0);
module_param_array(name, charp, NULL, 0);
MODULE_PARM_DESC(io, "WaveLAN I/O base address(es),required");
MODULE_PARM_DESC(irq, "WaveLAN IRQ number(s)");
MODULE_PARM_DESC(name, "WaveLAN interface neme(s)");
#endif /* MODULE */
#endif /* WAVELAN_P_H */