AcoSail/kernel_optimize_test
8
0
Fork 0
forked from luck/tmp_suning_uos_patched
Code Pull Requests Activity
74fd1b687f
kernel_optimize_test/include/linux/arcdevice.h
David Howells 7d12e780e0 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 15:10:12 +01:00

342 lines
12 KiB
C
Raw Blame History

/*
* INET An implementation of the TCP/IP protocol suite for the LINUX
* operating system. NET is implemented using the BSD Socket
* interface as the means of communication with the user level.
*
* Definitions used by the ARCnet driver.
*
* Authors: Avery Pennarun and David Woodhouse
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*
*/
#ifndef _LINUX_ARCDEVICE_H
#define _LINUX_ARCDEVICE_H
#include <asm/timex.h>
#include <linux/if_arcnet.h>
#ifdef __KERNEL__
#ifndef bool
#define bool int
#endif
/*
* RECON_THRESHOLD is the maximum number of RECON messages to receive
* within one minute before printing a "cabling problem" warning. The
* default value should be fine.
*
* After that, a "cabling restored" message will be printed on the next IRQ
* if no RECON messages have been received for 10 seconds.
*
* Do not define RECON_THRESHOLD at all if you want to disable this feature.
*/
#define RECON_THRESHOLD 30
/*
* Define this to the minimum "timeout" value. If a transmit takes longer
* than TX_TIMEOUT jiffies, Linux will abort the TX and retry. On a large
* network, or one with heavy network traffic, this timeout may need to be
* increased. The larger it is, though, the longer it will be between
* necessary transmits - don't set this too high.
*/
#define TX_TIMEOUT (HZ * 200 / 1000)
/* Display warnings about the driver being an ALPHA version. */
#undef ALPHA_WARNING
/*
* Debugging bitflags: each option can be enabled individually.
*
* Note: only debug flags included in the ARCNET_DEBUG_MAX define will
* actually be available. GCC will (at least, GCC 2.7.0 will) notice
* lines using a BUGLVL not in ARCNET_DEBUG_MAX and automatically optimize
* them out.
*/
#define D_NORMAL 1 /* important operational info */
#define D_EXTRA 2 /* useful, but non-vital information */
#define D_INIT 4 /* show init/probe messages */
#define D_INIT_REASONS 8 /* show reasons for discarding probes */
#define D_RECON 32 /* print a message whenever token is lost */
#define D_PROTO 64 /* debug auto-protocol support */
/* debug levels below give LOTS of output during normal operation! */
#define D_DURING 128 /* trace operations (including irq's) */
#define D_TX 256 /* show tx packets */
#define D_RX 512 /* show rx packets */
#define D_SKB 1024 /* show skb's */
#define D_SKB_SIZE 2048 /* show skb sizes */
#define D_TIMING 4096 /* show time needed to copy buffers to card */
#define D_DEBUG 8192 /* Very detailed debug line for line */
#ifndef ARCNET_DEBUG_MAX
#define ARCNET_DEBUG_MAX (127) /* change to ~0 if you want detailed debugging */
#endif
#ifndef ARCNET_DEBUG
#define ARCNET_DEBUG (D_NORMAL|D_EXTRA)
#endif
extern int arcnet_debug;
/* macros to simplify debug checking */
#define BUGLVL(x) if ((ARCNET_DEBUG_MAX)&arcnet_debug&(x))
#define BUGMSG2(x,msg,args...) do { BUGLVL(x) printk(msg, ## args); } while (0)
#define BUGMSG(x,msg,args...) \
BUGMSG2(x, "%s%6s: " msg, \
x==D_NORMAL ? KERN_WARNING \
: x < D_DURING ? KERN_INFO : KERN_DEBUG, \
dev->name , ## args)
/* see how long a function call takes to run, expressed in CPU cycles */
#define TIME(name, bytes, call) BUGLVL(D_TIMING) { \
unsigned long _x, _y; \
_x = get_cycles(); \
call; \
_y = get_cycles(); \
BUGMSG(D_TIMING, \
"%s: %d bytes in %lu cycles == " \
"%lu Kbytes/100Mcycle\n",\
name, bytes, _y - _x, \
100000000 / 1024 * bytes / (_y - _x + 1));\
} \
else { \
call;\
}
/*
* Time needed to reset the card - in ms (milliseconds). This works on my
* SMC PC100. I can't find a reference that tells me just how long I
* should wait.
*/
#define RESETtime (300)
/*
* These are the max/min lengths of packet payload, not including the
* arc_hardware header, but definitely including the soft header.
*
* Note: packet sizes 254, 255, 256 are impossible because of the way
* ARCnet registers work That's why RFC1201 defines "exception" packets.
* In non-RFC1201 protocols, we have to just tack some extra bytes on the
* end.
*/
#define MTU 253 /* normal packet max size */
#define MinTU 257 /* extended packet min size */
#define XMTU 508 /* extended packet max size */
/* status/interrupt mask bit fields */
#define TXFREEflag 0x01 /* transmitter available */
#define TXACKflag 0x02 /* transmitted msg. ackd */
#define RECONflag 0x04 /* network reconfigured */
#define TESTflag 0x08 /* test flag */
#define EXCNAKflag 0x08 /* excesive nak flag */
#define RESETflag 0x10 /* power-on-reset */
#define RES1flag 0x20 /* reserved - usually set by jumper */
#define RES2flag 0x40 /* reserved - usually set by jumper */
#define NORXflag 0x80 /* receiver inhibited */
/* Flags used for IO-mapped memory operations */
#define AUTOINCflag 0x40 /* Increase location with each access */
#define IOMAPflag 0x02 /* (for 90xx) Use IO mapped memory, not mmap */
#define ENABLE16flag 0x80 /* (for 90xx) Enable 16-bit mode */
/* in the command register, the following bits have these meanings:
* 0-2 command
* 3-4 page number (for enable rcv/xmt command)
* 7 receive broadcasts
*/
#define NOTXcmd 0x01 /* disable transmitter */
#define NORXcmd 0x02 /* disable receiver */
#define TXcmd 0x03 /* enable transmitter */
#define RXcmd 0x04 /* enable receiver */
#define CONFIGcmd 0x05 /* define configuration */
#define CFLAGScmd 0x06 /* clear flags */
#define TESTcmd 0x07 /* load test flags */
/* flags for "clear flags" command */
#define RESETclear 0x08 /* power-on-reset */
#define CONFIGclear 0x10 /* system reconfigured */
#define EXCNAKclear 0x0E /* Clear and acknowledge the excive nak bit */
/* flags for "load test flags" command */
#define TESTload 0x08 /* test flag (diagnostic) */
/* byte deposited into first address of buffers on reset */
#define TESTvalue 0321 /* that's octal for 0xD1 :) */
/* for "enable receiver" command */
#define RXbcasts 0x80 /* receive broadcasts */
/* flags for "define configuration" command */
#define NORMALconf 0x00 /* 1-249 byte packets */
#define EXTconf 0x08 /* 250-504 byte packets */
/* card feature flags, set during auto-detection.
* (currently only used by com20020pci)
*/
#define ARC_IS_5MBIT 1 /* card default speed is 5MBit */
#define ARC_CAN_10MBIT 2 /* card uses COM20022, supporting 10MBit,
but default is 2.5MBit. */
/* information needed to define an encapsulation driver */
struct ArcProto {
char suffix; /* a for RFC1201, e for ether-encap, etc. */
int mtu; /* largest possible packet */
int is_ip; /* This is a ip plugin - not a raw thing */
void (*rx) (struct net_device * dev, int bufnum,
struct archdr * pkthdr, int length);
int (*build_header) (struct sk_buff * skb, struct net_device *dev,
unsigned short ethproto, uint8_t daddr);
/* these functions return '1' if the skb can now be freed */
int (*prepare_tx) (struct net_device * dev, struct archdr * pkt, int length,
int bufnum);
int (*continue_tx) (struct net_device * dev, int bufnum);
int (*ack_tx) (struct net_device * dev, int acked);
};
extern struct ArcProto *arc_proto_map[256], *arc_proto_default,
*arc_bcast_proto, *arc_raw_proto;
/*
* "Incoming" is information needed for each address that could be sending
* to us. Mostly for partially-received split packets.
*/
struct Incoming {
struct sk_buff *skb; /* packet data buffer */
uint16_t sequence; /* sequence number of assembly */
uint8_t lastpacket, /* number of last packet (from 1) */
numpackets; /* number of packets in split */
};
/* only needed for RFC1201 */
struct Outgoing {
struct ArcProto *proto; /* protocol driver that owns this:
* if NULL, no packet is pending.
*/
struct sk_buff *skb; /* buffer from upper levels */
struct archdr *pkt; /* a pointer into the skb */
uint16_t length, /* bytes total */
dataleft, /* bytes left */
segnum, /* segment being sent */
numsegs; /* number of segments */
};
struct arcnet_local {
struct net_device_stats stats;
uint8_t config, /* current value of CONFIG register */
timeout, /* Extended timeout for COM20020 */
backplane, /* Backplane flag for COM20020 */
clockp, /* COM20020 clock divider */
clockm, /* COM20020 clock multiplier flag */
setup, /* Contents of setup1 register */
setup2, /* Contents of setup2 register */
intmask; /* current value of INTMASK register */
uint8_t default_proto[256]; /* default encap to use for each host */
int cur_tx, /* buffer used by current transmit, or -1 */
next_tx, /* buffer where a packet is ready to send */
cur_rx; /* current receive buffer */
int lastload_dest, /* can last loaded packet be acked? */
lasttrans_dest; /* can last TX'd packet be acked? */
int timed_out; /* need to process TX timeout and drop packet */
unsigned long last_timeout; /* time of last reported timeout */
char *card_name; /* card ident string */
int card_flags; /* special card features */
/* On preemtive and SMB a lock is needed */
spinlock_t lock;
/*
* Buffer management: an ARCnet card has 4 x 512-byte buffers, each of
* which can be used for either sending or receiving. The new dynamic
* buffer management routines use a simple circular queue of available
* buffers, and take them as they're needed. This way, we simplify
* situations in which we (for example) want to pre-load a transmit
* buffer, or start receiving while we copy a received packet to
* memory.
*
* The rules: only the interrupt handler is allowed to _add_ buffers to
* the queue; thus, this doesn't require a lock. Both the interrupt
* handler and the transmit function will want to _remove_ buffers, so
* we need to handle the situation where they try to do it at the same
* time.
*
* If next_buf == first_free_buf, the queue is empty. Since there are
* only four possible buffers, the queue should never be full.
*/
atomic_t buf_lock;
int buf_queue[5];
int next_buf, first_free_buf;
/* network "reconfiguration" handling */
time_t first_recon, /* time of "first" RECON message to count */
last_recon; /* time of most recent RECON */
int num_recons; /* number of RECONs between first and last. */
bool network_down; /* do we think the network is down? */
bool excnak_pending; /* We just got an excesive nak interrupt */
struct {
uint16_t sequence; /* sequence number (incs with each packet) */
uint16_t aborted_seq;
struct Incoming incoming[256]; /* one from each address */
} rfc1201;
/* really only used by rfc1201, but we'll pretend it's not */
struct Outgoing outgoing; /* packet currently being sent */
/* hardware-specific functions */
struct {
struct module *owner;
void (*command) (struct net_device * dev, int cmd);
int (*status) (struct net_device * dev);
void (*intmask) (struct net_device * dev, int mask);
bool (*reset) (struct net_device * dev, bool really_reset);
void (*open) (struct net_device * dev);
void (*close) (struct net_device * dev);
void (*copy_to_card) (struct net_device * dev, int bufnum, int offset,
void *buf, int count);
void (*copy_from_card) (struct net_device * dev, int bufnum, int offset,
void *buf, int count);
} hw;
void __iomem *mem_start; /* pointer to ioremap'ed MMIO */
};
#define ARCRESET(x) (lp->hw.reset(dev, (x)))
#define ACOMMAND(x) (lp->hw.command(dev, (x)))
#define ASTATUS() (lp->hw.status(dev))
#define AINTMASK(x) (lp->hw.intmask(dev, (x)))
#if ARCNET_DEBUG_MAX & D_SKB
void arcnet_dump_skb(struct net_device *dev, struct sk_buff *skb, char *desc);
#else
#define arcnet_dump_skb(dev,skb,desc) ;
#endif
void arcnet_unregister_proto(struct ArcProto *proto);
irqreturn_t arcnet_interrupt(int irq, void *dev_id);
struct net_device *alloc_arcdev(char *name);
#endif /* __KERNEL__ */
#endif /* _LINUX_ARCDEVICE_H */
View Git Blame Copy Permalink