linux/drivers/net/tokenring/ibmtr.c

1965 lines
61 KiB
C
Raw Normal View History

/* ibmtr.c: A shared-memory IBM Token Ring 16/4 driver for linux
*
* Written 1993 by Mark Swanson and Peter De Schrijver.
* This software may be used and distributed according to the terms
* of the GNU General Public License, incorporated herein by reference.
*
* This device driver should work with Any IBM Token Ring Card that does
* not use DMA.
*
* I used Donald Becker's (becker@scyld.com) device driver work
* as a base for most of my initial work.
*
* Changes by Peter De Schrijver
* (Peter.Deschrijver@linux.cc.kuleuven.ac.be) :
*
* + changed name to ibmtr.c in anticipation of other tr boards.
* + changed reset code and adapter open code.
* + added SAP open code.
* + a first attempt to write interrupt, transmit and receive routines.
*
* Changes by David W. Morris (dwm@shell.portal.com) :
* 941003 dwm: - Restructure tok_probe for multiple adapters, devices.
* + Add comments, misc reorg for clarity.
* + Flatten interrupt handler levels.
*
* Changes by Farzad Farid (farzy@zen.via.ecp.fr)
* and Pascal Andre (andre@chimay.via.ecp.fr) (March 9 1995) :
* + multi ring support clean up.
* + RFC1042 compliance enhanced.
*
* Changes by Pascal Andre (andre@chimay.via.ecp.fr) (September 7 1995) :
* + bug correction in tr_tx
* + removed redundant information display
* + some code reworking
*
* Changes by Michel Lespinasse (walken@via.ecp.fr),
* Yann Doussot (doussot@via.ecp.fr) and Pascal Andre (andre@via.ecp.fr)
* (February 18, 1996) :
* + modified shared memory and mmio access port the driver to
* alpha platform (structure access -> readb/writeb)
*
* Changes by Steve Kipisz (bungy@ibm.net or kipisz@vnet.ibm.com)
* (January 18 1996):
* + swapped WWOR and WWCR in ibmtr.h
* + moved some init code from tok_probe into trdev_init. The
* PCMCIA code can call trdev_init to complete initializing
* the driver.
* + added -DPCMCIA to support PCMCIA
* + detecting PCMCIA Card Removal in interrupt handler. If
* ISRP is FF, then a PCMCIA card has been removed
* 10/2000 Burt needed a new method to avoid crashing the OS
*
* Changes by Paul Norton (pnorton@cts.com) :
* + restructured the READ.LOG logic to prevent the transmit SRB
* from being rudely overwritten before the transmit cycle is
* complete. (August 15 1996)
* + completed multiple adapter support. (November 20 1996)
* + implemented csum_partial_copy in tr_rx and increased receive
* buffer size and count. Minor fixes. (March 15, 1997)
*
* Changes by Christopher Turcksin <wabbit@rtfc.demon.co.uk>
* + Now compiles ok as a module again.
*
* Changes by Paul Norton (pnorton@ieee.org) :
* + moved the header manipulation code in tr_tx and tr_rx to
* net/802/tr.c. (July 12 1997)
* + add retry and timeout on open if cable disconnected. (May 5 1998)
* + lifted 2000 byte mtu limit. now depends on shared-RAM size.
* May 25 1998)
* + can't allocate 2k recv buff at 8k shared-RAM. (20 October 1998)
*
* Changes by Joel Sloan (jjs@c-me.com) :
* + disable verbose debug messages by default - to enable verbose
* debugging, edit the IBMTR_DEBUG_MESSAGES define below
*
* Changes by Mike Phillips <phillim@amtrak.com> :
* + Added extra #ifdef's to work with new PCMCIA Token Ring Code.
* The PCMCIA code now just sets up the card so it can be recognized
* by ibmtr_probe. Also checks allocated memory vs. on-board memory
* for correct figure to use.
*
* Changes by Tim Hockin (thockin@isunix.it.ilstu.edu) :
* + added spinlocks for SMP sanity (10 March 1999)
*
* Changes by Jochen Friedrich to enable RFC1469 Option 2 multicasting
* i.e. using functional address C0 00 00 04 00 00 to transmit and
* receive multicast packets.
*
* Changes by Mike Sullivan (based on original sram patch by Dave Grothe
* to support windowing into on adapter shared ram.
* i.e. Use LANAID to setup a PnP configuration with 16K RAM. Paging
* will shift this 16K window over the entire available shared RAM.
*
* Changes by Peter De Schrijver (p2@mind.be) :
* + fixed a problem with PCMCIA card removal
*
* Change by Mike Sullivan et al.:
* + added turbo card support. No need to use lanaid to configure
* the adapter into isa compatibility mode.
*
* Changes by Burt Silverman to allow the computer to behave nicely when
* a cable is pulled or not in place, or a PCMCIA card is removed hot.
*/
/* change the define of IBMTR_DEBUG_MESSAGES to a nonzero value
in the event that chatty debug messages are desired - jjs 12/30/98 */
#define IBMTR_DEBUG_MESSAGES 0
#include <linux/module.h>
#include <linux/sched.h>
#ifdef PCMCIA /* required for ibmtr_cs.c to build */
#undef MODULE /* yes, really */
#undef ENABLE_PAGING
#else
#define ENABLE_PAGING 1
#endif
/* changes the output format of driver initialization */
#define TR_VERBOSE 0
/* some 95 OS send many non UI frame; this allow removing the warning */
#define TR_FILTERNONUI 1
#include <linux/interrupt.h>
#include <linux/ioport.h>
#include <linux/netdevice.h>
#include <linux/ip.h>
#include <linux/trdevice.h>
#include <linux/ibmtr.h>
#include <net/checksum.h>
#include <asm/io.h>
#define DPRINTK(format, args...) printk("%s: " format, dev->name , ## args)
#define DPRINTD(format, args...) DummyCall("%s: " format, dev->name , ## args)
/* version and credits */
#ifndef PCMCIA
static char version[] __devinitdata =
"\nibmtr.c: v1.3.57 8/ 7/94 Peter De Schrijver and Mark Swanson\n"
" v2.1.125 10/20/98 Paul Norton <pnorton@ieee.org>\n"
" v2.2.0 12/30/98 Joel Sloan <jjs@c-me.com>\n"
" v2.2.1 02/08/00 Mike Sullivan <sullivam@us.ibm.com>\n"
" v2.2.2 07/27/00 Burt Silverman <burts@us.ibm.com>\n"
" v2.4.0 03/01/01 Mike Sullivan <sullivan@us.ibm.com>\n";
#endif
/* this allows displaying full adapter information */
static char *channel_def[] __devinitdata = { "ISA", "MCA", "ISA P&P" };
static char pcchannelid[] __devinitdata = {
0x05, 0x00, 0x04, 0x09,
0x04, 0x03, 0x04, 0x0f,
0x03, 0x06, 0x03, 0x01,
0x03, 0x01, 0x03, 0x00,
0x03, 0x09, 0x03, 0x09,
0x03, 0x00, 0x02, 0x00
};
static char mcchannelid[] __devinitdata = {
0x04, 0x0d, 0x04, 0x01,
0x05, 0x02, 0x05, 0x03,
0x03, 0x06, 0x03, 0x03,
0x05, 0x08, 0x03, 0x04,
0x03, 0x05, 0x03, 0x01,
0x03, 0x08, 0x02, 0x00
};
static char __devinit *adapter_def(char type)
{
switch (type) {
case 0xF: return "PC Adapter | PC Adapter II | Adapter/A";
case 0xE: return "16/4 Adapter | 16/4 Adapter/A (long)";
case 0xD: return "16/4 Adapter/A (short) | 16/4 ISA-16 Adapter";
case 0xC: return "Auto 16/4 Adapter";
default: return "adapter (unknown type)";
}
};
#define TRC_INIT 0x01 /* Trace initialization & PROBEs */
#define TRC_INITV 0x02 /* verbose init trace points */
static unsigned char ibmtr_debug_trace = 0;
static int ibmtr_probe1(struct net_device *dev, int ioaddr);
static unsigned char get_sram_size(struct tok_info *adapt_info);
static int trdev_init(struct net_device *dev);
static int tok_open(struct net_device *dev);
static int tok_init_card(struct net_device *dev);
static void tok_open_adapter(unsigned long dev_addr);
static void open_sap(unsigned char type, struct net_device *dev);
static void tok_set_multicast_list(struct net_device *dev);
static netdev_tx_t tok_send_packet(struct sk_buff *skb,
struct net_device *dev);
static int tok_close(struct net_device *dev);
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 15:55:46 +02:00
static irqreturn_t tok_interrupt(int irq, void *dev_id);
static void initial_tok_int(struct net_device *dev);
static void tr_tx(struct net_device *dev);
static void tr_rx(struct net_device *dev);
static void ibmtr_reset_timer(struct timer_list*tmr,struct net_device *dev);
static void tok_rerun(unsigned long dev_addr);
static void ibmtr_readlog(struct net_device *dev);
static int ibmtr_change_mtu(struct net_device *dev, int mtu);
static void find_turbo_adapters(int *iolist);
static int ibmtr_portlist[IBMTR_MAX_ADAPTERS+1] __devinitdata = {
0xa20, 0xa24, 0, 0, 0
};
static int __devinitdata turbo_io[IBMTR_MAX_ADAPTERS] = {0};
static int __devinitdata turbo_irq[IBMTR_MAX_ADAPTERS] = {0};
static int __devinitdata turbo_searched = 0;
#ifndef PCMCIA
static __u32 ibmtr_mem_base __devinitdata = 0xd0000;
#endif
static void __devinit PrtChanID(char *pcid, short stride)
{
short i, j;
for (i = 0, j = 0; i < 24; i++, j += stride)
printk("%1x", ((int) pcid[j]) & 0x0f);
printk("\n");
}
static void __devinit HWPrtChanID(void __iomem *pcid, short stride)
{
short i, j;
for (i = 0, j = 0; i < 24; i++, j += stride)
printk("%1x", ((int) readb(pcid + j)) & 0x0f);
printk("\n");
}
/* We have to ioremap every checked address, because isa_readb is
* going away.
*/
static void __devinit find_turbo_adapters(int *iolist)
{
int ram_addr;
int index=0;
void __iomem *chanid;
int found_turbo=0;
unsigned char *tchanid, ctemp;
int i, j;
unsigned long jif;
void __iomem *ram_mapped ;
if (turbo_searched == 1) return;
turbo_searched=1;
for (ram_addr=0xC0000; ram_addr < 0xE0000; ram_addr+=0x2000) {
__u32 intf_tbl=0;
found_turbo=1;
ram_mapped = ioremap((u32)ram_addr,0x1fff) ;
if (ram_mapped==NULL)
continue ;
chanid=(CHANNEL_ID + ram_mapped);
tchanid=pcchannelid;
ctemp=readb(chanid) & 0x0f;
if (ctemp != *tchanid) continue;
for (i=2,j=1; i<=46; i=i+2,j++) {
if ((readb(chanid+i) & 0x0f) != tchanid[j]){
found_turbo=0;
break;
}
}
if (!found_turbo) continue;
writeb(0x90, ram_mapped+0x1E01);
for(i=2; i<0x0f; i++) {
writeb(0x00, ram_mapped+0x1E01+i);
}
writeb(0x00, ram_mapped+0x1E01);
for(jif=jiffies+TR_BUSY_INTERVAL; time_before_eq(jiffies,jif););
intf_tbl=ntohs(readw(ram_mapped+ACA_OFFSET+ACA_RW+WRBR_EVEN));
if (intf_tbl) {
#if IBMTR_DEBUG_MESSAGES
printk("ibmtr::find_turbo_adapters, Turbo found at "
"ram_addr %x\n",ram_addr);
printk("ibmtr::find_turbo_adapters, interface_table ");
for(i=0; i<6; i++) {
printk("%x:",readb(ram_addr+intf_tbl+i));
}
printk("\n");
#endif
turbo_io[index]=ntohs(readw(ram_mapped+intf_tbl+4));
turbo_irq[index]=readb(ram_mapped+intf_tbl+3);
outb(0, turbo_io[index] + ADAPTRESET);
for(jif=jiffies+TR_RST_TIME;time_before_eq(jiffies,jif););
outb(0, turbo_io[index] + ADAPTRESETREL);
index++;
continue;
}
#if IBMTR_DEBUG_MESSAGES
printk("ibmtr::find_turbo_adapters, ibmtr card found at"
" %x but not a Turbo model\n",ram_addr);
#endif
iounmap(ram_mapped) ;
} /* for */
for(i=0; i<IBMTR_MAX_ADAPTERS; i++) {
if(!turbo_io[i]) break;
for (j=0; j<IBMTR_MAX_ADAPTERS; j++) {
if ( iolist[j] && iolist[j] != turbo_io[i]) continue;
iolist[j]=turbo_io[i];
break;
}
}
}
static void ibmtr_cleanup_card(struct net_device *dev)
{
if (dev->base_addr) {
outb(0,dev->base_addr+ADAPTRESET);
schedule_timeout_uninterruptible(TR_RST_TIME); /* wait 50ms */
outb(0,dev->base_addr+ADAPTRESETREL);
}
#ifndef PCMCIA
free_irq(dev->irq, dev);
release_region(dev->base_addr, IBMTR_IO_EXTENT);
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
iounmap(ti->mmio);
iounmap(ti->sram_virt);
}
#endif
}
/****************************************************************************
* ibmtr_probe(): Routine specified in the network device structure
* to probe for an IBM Token Ring Adapter. Routine outline:
* I. Interrogate hardware to determine if an adapter exists
* and what the speeds and feeds are
* II. Setup data structures to control execution based upon
* adapter characteristics.
*
* We expect ibmtr_probe to be called once for each device entry
* which references it.
****************************************************************************/
static int __devinit ibmtr_probe(struct net_device *dev)
{
int i;
int base_addr = dev->base_addr;
if (base_addr && base_addr <= 0x1ff) /* Don't probe at all. */
return -ENXIO;
if (base_addr > 0x1ff) { /* Check a single specified location. */
if (!ibmtr_probe1(dev, base_addr)) return 0;
return -ENODEV;
}
find_turbo_adapters(ibmtr_portlist);
for (i = 0; ibmtr_portlist[i]; i++) {
int ioaddr = ibmtr_portlist[i];
if (!ibmtr_probe1(dev, ioaddr)) return 0;
}
return -ENODEV;
}
int __devinit ibmtr_probe_card(struct net_device *dev)
ibmtr: Drain rich supply of modpost warnings. Building ibmtr as a module produces a spectacular pile of modpost warnings: WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x450) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x454) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x458) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x45c) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4e8) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4ec) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4f0) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x4f4) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x500) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x504) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x508) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:find_turbo_adapters from .text between 'ibmtr_probe_card' (at offset 0x50c) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x520) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x524) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x528) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data:ibmtr_portlist from .text between 'ibmtr_probe_card' (at offset 0x534) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x540) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x544) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x548) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.data: from .text between 'ibmtr_probe_card' (at offset 0x54c) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x558) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x55c) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x560) and 'ibmtr_reset_timer' WARNING: drivers/net/tokenring/ibmtr.o - Section mismatch: reference to .init.text:ibmtr_probe1 from .text between 'ibmtr_probe_card' (at offset 0x564) and 'ibmtr_reset_timer' Fix by making ibmtr_probe an __init function. While at it, move move ibmtr_probe_card below ibmtr_probe so the protoype for ibmtr_probe can be deleted. Signed-off-by: Ralf Baechle <ralf@linux-mips.org> drivers/net/tokenring/ibmtr.c | 25 ++++++++++++------------- 1 file changed, 12 insertions(+), 13 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-03-08 13:53:07 +01:00
{
int err = ibmtr_probe(dev);
if (!err) {
err = register_netdev(dev);
if (err)
ibmtr_cleanup_card(dev);
}
return err;
}
/*****************************************************************************/
static int __devinit ibmtr_probe1(struct net_device *dev, int PIOaddr)
{
unsigned char segment, intr=0, irq=0, i, j, cardpresent=NOTOK, temp=0;
void __iomem * t_mmio = NULL;
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
void __iomem *cd_chanid;
unsigned char *tchanid, ctemp;
#ifndef PCMCIA
unsigned char t_irq=0;
unsigned long timeout;
static int version_printed;
#endif
/* Query the adapter PIO base port which will return
* indication of where MMIO was placed. We also have a
* coded interrupt number.
*/
segment = inb(PIOaddr);
if (segment < 0x40 || segment > 0xe0) {
/* Out of range values so we'll assume non-existent IO device
* but this is not necessarily a problem, esp if a turbo
* adapter is being used. */
#if IBMTR_DEBUG_MESSAGES
DPRINTK("ibmtr_probe1(): unhappy that inb(0x%X) == 0x%X, "
"Hardware Problem?\n",PIOaddr,segment);
#endif
return -ENODEV;
}
/*
* Compute the linear base address of the MMIO area
* as LINUX doesn't care about segments
*/
t_mmio = ioremap(((__u32) (segment & 0xfc) << 11) + 0x80000,2048);
if (!t_mmio) {
DPRINTK("Cannot remap mmiobase memory area") ;
return -ENODEV ;
}
intr = segment & 0x03; /* low bits is coded interrupt # */
if (ibmtr_debug_trace & TRC_INIT)
DPRINTK("PIOaddr: %4hx seg/intr: %2x mmio base: %p intr: %d\n"
, PIOaddr, (int) segment, t_mmio, (int) intr);
/*
* Now we will compare expected 'channelid' strings with
* what we is there to learn of ISA/MCA or not TR card
*/
#ifdef PCMCIA
iounmap(t_mmio);
t_mmio = ti->mmio; /*BMS to get virtual address */
irq = ti->irq; /*BMS to display the irq! */
#endif
cd_chanid = (CHANNEL_ID + t_mmio); /* for efficiency */
tchanid = pcchannelid;
cardpresent = TR_ISA; /* try ISA */
/* Suboptimize knowing first byte different */
ctemp = readb(cd_chanid) & 0x0f;
if (ctemp != *tchanid) { /* NOT ISA card, try MCA */
tchanid = mcchannelid;
cardpresent = TR_MCA;
if (ctemp != *tchanid) /* Neither ISA nor MCA */
cardpresent = NOTOK;
}
if (cardpresent != NOTOK) {
/* Know presumed type, try rest of ID */
for (i = 2, j = 1; i <= 46; i = i + 2, j++) {
if( (readb(cd_chanid+i)&0x0f) == tchanid[j]) continue;
/* match failed, not TR card */
cardpresent = NOTOK;
break;
}
}
/*
* If we have an ISA board check for the ISA P&P version,
* as it has different IRQ settings
*/
if (cardpresent == TR_ISA && (readb(AIPFID + t_mmio) == 0x0e))
cardpresent = TR_ISAPNP;
if (cardpresent == NOTOK) { /* "channel_id" did not match, report */
if (!(ibmtr_debug_trace & TRC_INIT)) {
#ifndef PCMCIA
iounmap(t_mmio);
#endif
return -ENODEV;
}
DPRINTK( "Channel ID string not found for PIOaddr: %4hx\n",
PIOaddr);
DPRINTK("Expected for ISA: ");
PrtChanID(pcchannelid, 1);
DPRINTK(" found: ");
/* BMS Note that this can be misleading, when hardware is flaky, because you
are reading it a second time here. So with my flaky hardware, I'll see my-
self in this block, with the HW ID matching the ISA ID exactly! */
HWPrtChanID(cd_chanid, 2);
DPRINTK("Expected for MCA: ");
PrtChanID(mcchannelid, 1);
}
/* Now, setup some of the pl0 buffers for this driver.. */
/* If called from PCMCIA, it is already set up, so no need to
waste the memory, just use the existing structure */
#ifndef PCMCIA
ti->mmio = t_mmio;
for (i = 0; i < IBMTR_MAX_ADAPTERS; i++) {
if (turbo_io[i] != PIOaddr)
continue;
#if IBMTR_DEBUG_MESSAGES
printk("ibmtr::tr_probe1, setting PIOaddr %x to Turbo\n",
PIOaddr);
#endif
ti->turbo = 1;
t_irq = turbo_irq[i];
}
#endif /* !PCMCIA */
ti->readlog_pending = 0;
init_waitqueue_head(&ti->wait_for_reset);
/* if PCMCIA, the card can be recognized as either TR_ISA or TR_ISAPNP
* depending which card is inserted. */
#ifndef PCMCIA
switch (cardpresent) {
case TR_ISA:
if (intr == 0) irq = 9; /* irq2 really is irq9 */
if (intr == 1) irq = 3;
if (intr == 2) irq = 6;
if (intr == 3) irq = 7;
ti->adapter_int_enable = PIOaddr + ADAPTINTREL;
break;
case TR_MCA:
if (intr == 0) irq = 9;
if (intr == 1) irq = 3;
if (intr == 2) irq = 10;
if (intr == 3) irq = 11;
ti->global_int_enable = 0;
ti->adapter_int_enable = 0;
ti->sram_phys=(__u32)(inb(PIOaddr+ADAPTRESETREL) & 0xfe) << 12;
break;
case TR_ISAPNP:
if (!t_irq) {
if (intr == 0) irq = 9;
if (intr == 1) irq = 3;
if (intr == 2) irq = 10;
if (intr == 3) irq = 11;
} else
irq=t_irq;
timeout = jiffies + TR_SPIN_INTERVAL;
while (!readb(ti->mmio + ACA_OFFSET + ACA_RW + RRR_EVEN)){
if (!time_after(jiffies, timeout)) continue;
DPRINTK( "Hardware timeout during initialization.\n");
iounmap(t_mmio);
return -ENODEV;
}
ti->sram_phys =
((__u32)readb(ti->mmio+ACA_OFFSET+ACA_RW+RRR_EVEN)<<12);
ti->adapter_int_enable = PIOaddr + ADAPTINTREL;
break;
} /*end switch (cardpresent) */
#endif /*not PCMCIA */
if (ibmtr_debug_trace & TRC_INIT) { /* just report int */
DPRINTK("irq=%d", irq);
printk(", sram_phys=0x%x", ti->sram_phys);
if(ibmtr_debug_trace&TRC_INITV){ /* full chat in verbose only */
DPRINTK(", ti->mmio=%p", ti->mmio);
printk(", segment=%02X", segment);
}
printk(".\n");
}
/* Get hw address of token ring card */
j = 0;
for (i = 0; i < 0x18; i = i + 2) {
/* technical reference states to do this */
temp = readb(ti->mmio + AIP + i) & 0x0f;
ti->hw_address[j] = temp;
if (j & 1)
dev->dev_addr[(j / 2)] =
ti->hw_address[j]+ (ti->hw_address[j - 1] << 4);
++j;
}
/* get Adapter type: 'F' = Adapter/A, 'E' = 16/4 Adapter II,... */
ti->adapter_type = readb(ti->mmio + AIPADAPTYPE);
/* get Data Rate: F=4Mb, E=16Mb, D=4Mb & 16Mb ?? */
ti->data_rate = readb(ti->mmio + AIPDATARATE);
/* Get Early Token Release support?: F=no, E=4Mb, D=16Mb, C=4&16Mb */
ti->token_release = readb(ti->mmio + AIPEARLYTOKEN);
/* How much shared RAM is on adapter ? */
if (ti->turbo) {
ti->avail_shared_ram=127;
} else {
ti->avail_shared_ram = get_sram_size(ti);/*in 512 byte units */
}
/* We need to set or do a bunch of work here based on previous results*/
/* Support paging? What sizes?: F=no, E=16k, D=32k, C=16 & 32k */
ti->shared_ram_paging = readb(ti->mmio + AIPSHRAMPAGE);
/* Available DHB 4Mb size: F=2048, E=4096, D=4464 */
switch (readb(ti->mmio + AIP4MBDHB)) {
case 0xe: ti->dhb_size4mb = 4096; break;
case 0xd: ti->dhb_size4mb = 4464; break;
default: ti->dhb_size4mb = 2048; break;
}
/* Available DHB 16Mb size: F=2048, E=4096, D=8192, C=16384, B=17960 */
switch (readb(ti->mmio + AIP16MBDHB)) {
case 0xe: ti->dhb_size16mb = 4096; break;
case 0xd: ti->dhb_size16mb = 8192; break;
case 0xc: ti->dhb_size16mb = 16384; break;
case 0xb: ti->dhb_size16mb = 17960; break;
default: ti->dhb_size16mb = 2048; break;
}
/* We must figure out how much shared memory space this adapter
* will occupy so that if there are two adapters we can fit both
* in. Given a choice, we will limit this adapter to 32K. The
* maximum space will will use for two adapters is 64K so if the
* adapter we are working on demands 64K (it also doesn't support
* paging), then only one adapter can be supported.
*/
/*
* determine how much of total RAM is mapped into PC space
*/
ti->mapped_ram_size= /*sixteen to onehundredtwentyeight 512byte blocks*/
1<< ((readb(ti->mmio+ACA_OFFSET+ACA_RW+RRR_ODD) >> 2 & 0x03) + 4);
ti->page_mask = 0;
if (ti->turbo) ti->page_mask=0xf0;
else if (ti->shared_ram_paging == 0xf); /* No paging in adapter */
else {
#ifdef ENABLE_PAGING
unsigned char pg_size = 0;
/* BMS: page size: PCMCIA, use configuration register;
ISAPNP, use LANAIDC config tool from www.ibm.com */
switch (ti->shared_ram_paging) {
case 0xf:
break;
case 0xe:
ti->page_mask = (ti->mapped_ram_size == 32) ? 0xc0 : 0;
pg_size = 32; /* 16KB page size */
break;
case 0xd:
ti->page_mask = (ti->mapped_ram_size == 64) ? 0x80 : 0;
pg_size = 64; /* 32KB page size */
break;
case 0xc:
switch (ti->mapped_ram_size) {
case 32:
ti->page_mask = 0xc0;
pg_size = 32;
break;
case 64:
ti->page_mask = 0x80;
pg_size = 64;
break;
}
break;
default:
DPRINTK("Unknown shared ram paging info %01X\n",
ti->shared_ram_paging);
iounmap(t_mmio);
return -ENODEV;
break;
} /*end switch shared_ram_paging */
if (ibmtr_debug_trace & TRC_INIT)
DPRINTK("Shared RAM paging code: %02X, "
"mapped RAM size: %dK, shared RAM size: %dK, "
"page mask: %02X\n:",
ti->shared_ram_paging, ti->mapped_ram_size / 2,
ti->avail_shared_ram / 2, ti->page_mask);
#endif /*ENABLE_PAGING */
}
#ifndef PCMCIA
/* finish figuring the shared RAM address */
if (cardpresent == TR_ISA) {
static const __u32 ram_bndry_mask[] = {
0xffffe000, 0xffffc000, 0xffff8000, 0xffff0000
};
__u32 new_base, rrr_32, chk_base, rbm;
rrr_32=readb(ti->mmio+ACA_OFFSET+ACA_RW+RRR_ODD) >> 2 & 0x03;
rbm = ram_bndry_mask[rrr_32];
new_base = (ibmtr_mem_base + (~rbm)) & rbm;/* up to boundary */
chk_base = new_base + (ti->mapped_ram_size << 9);
if (chk_base > (ibmtr_mem_base + IBMTR_SHARED_RAM_SIZE)) {
DPRINTK("Shared RAM for this adapter (%05x) exceeds "
"driver limit (%05x), adapter not started.\n",
chk_base, ibmtr_mem_base + IBMTR_SHARED_RAM_SIZE);
iounmap(t_mmio);
return -ENODEV;
} else { /* seems cool, record what we have figured out */
ti->sram_base = new_base >> 12;
ibmtr_mem_base = chk_base;
}
}
else ti->sram_base = ti->sram_phys >> 12;
/* The PCMCIA has already got the interrupt line and the io port,
so no chance of anybody else getting it - MLP */
if (request_irq(dev->irq = irq, tok_interrupt, 0, "ibmtr", dev) != 0) {
DPRINTK("Could not grab irq %d. Halting Token Ring driver.\n",
irq);
iounmap(t_mmio);
return -ENODEV;
}
/*?? Now, allocate some of the PIO PORTs for this driver.. */
/* record PIOaddr range as busy */
if (!request_region(PIOaddr, IBMTR_IO_EXTENT, "ibmtr")) {
DPRINTK("Could not grab PIO range. Halting driver.\n");
free_irq(dev->irq, dev);
iounmap(t_mmio);
return -EBUSY;
}
if (!version_printed++) {
printk(version);
}
#endif /* !PCMCIA */
DPRINTK("%s %s found\n",
channel_def[cardpresent - 1], adapter_def(ti->adapter_type));
DPRINTK("using irq %d, PIOaddr %hx, %dK shared RAM.\n",
irq, PIOaddr, ti->mapped_ram_size / 2);
DPRINTK("Hardware address : %pM\n", dev->dev_addr);
if (ti->page_mask)
DPRINTK("Shared RAM paging enabled. "
"Page size: %uK Shared Ram size %dK\n",
((ti->page_mask^0xff)+1) >>2, ti->avail_shared_ram / 2);
else
DPRINTK("Shared RAM paging disabled. ti->page_mask %x\n",
ti->page_mask);
/* Calculate the maximum DHB we can use */
/* two cases where avail_shared_ram doesn't equal mapped_ram_size:
1. avail_shared_ram is 127 but mapped_ram_size is 128 (typical)
2. user has configured adapter for less than avail_shared_ram
but is not using paging (she should use paging, I believe)
*/
if (!ti->page_mask) {
ti->avail_shared_ram=
min(ti->mapped_ram_size,ti->avail_shared_ram);
}
switch (ti->avail_shared_ram) {
case 16: /* 8KB shared RAM */
ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)2048);
ti->rbuf_len4 = 1032;
ti->rbuf_cnt4=2;
ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)2048);
ti->rbuf_len16 = 1032;
ti->rbuf_cnt16=2;
break;
case 32: /* 16KB shared RAM */
ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464);
ti->rbuf_len4 = 1032;
ti->rbuf_cnt4=4;
ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)4096);
ti->rbuf_len16 = 1032; /*1024 usable */
ti->rbuf_cnt16=4;
break;
case 64: /* 32KB shared RAM */
ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464);
ti->rbuf_len4 = 1032;
ti->rbuf_cnt4=6;
ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)10240);
ti->rbuf_len16 = 1032;
ti->rbuf_cnt16=6;
break;
case 127: /* 63.5KB shared RAM */
ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464);
ti->rbuf_len4 = 1032;
ti->rbuf_cnt4=6;
ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)16384);
ti->rbuf_len16 = 1032;
ti->rbuf_cnt16=16;
break;
case 128: /* 64KB shared RAM */
ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464);
ti->rbuf_len4 = 1032;
ti->rbuf_cnt4=6;
ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)17960);
ti->rbuf_len16 = 1032;
ti->rbuf_cnt16=16;
break;
default:
ti->dhb_size4mb = 2048;
ti->rbuf_len4 = 1032;
ti->rbuf_cnt4=2;
ti->dhb_size16mb = 2048;
ti->rbuf_len16 = 1032;
ti->rbuf_cnt16=2;
break;
}
/* this formula is not smart enough for the paging case
ti->rbuf_cnt<x> = (ti->avail_shared_ram * BLOCKSZ - ADAPT_PRIVATE -
ARBLENGTH - SSBLENGTH - DLC_MAX_SAP * SAPLENGTH -
DLC_MAX_STA * STALENGTH - ti->dhb_size<x>mb * NUM_DHB -
SRBLENGTH - ASBLENGTH) / ti->rbuf_len<x>;
*/
ti->maxmtu16 = (ti->rbuf_len16 - 8) * ti->rbuf_cnt16 - TR_HLEN;
ti->maxmtu4 = (ti->rbuf_len4 - 8) * ti->rbuf_cnt4 - TR_HLEN;
/*BMS assuming 18 bytes of Routing Information (usually works) */
DPRINTK("Maximum Receive Internet Protocol MTU 16Mbps: %d, 4Mbps: %d\n",
ti->maxmtu16, ti->maxmtu4);
dev->base_addr = PIOaddr; /* set the value for device */
dev->mem_start = ti->sram_base << 12;
dev->mem_end = dev->mem_start + (ti->mapped_ram_size << 9) - 1;
trdev_init(dev);
return 0; /* Return 0 to indicate we have found a Token Ring card. */
} /*ibmtr_probe1() */
/*****************************************************************************/
/* query the adapter for the size of shared RAM */
/* the function returns the RAM size in units of 512 bytes */
static unsigned char __devinit get_sram_size(struct tok_info *adapt_info)
{
unsigned char avail_sram_code;
static unsigned char size_code[] = { 0, 16, 32, 64, 127, 128 };
/* Adapter gives
'F' -- use RRR bits 3,2
'E' -- 8kb 'D' -- 16kb
'C' -- 32kb 'A' -- 64KB
'B' - 64KB less 512 bytes at top
(WARNING ... must zero top bytes in INIT */
avail_sram_code = 0xf - readb(adapt_info->mmio + AIPAVAILSHRAM);
if (avail_sram_code) return size_code[avail_sram_code];
else /* for code 'F', must compute size from RRR(3,2) bits */
return 1 <<
((readb(adapt_info->mmio+ACA_OFFSET+ACA_RW+RRR_ODD)>>2&3)+4);
}
/*****************************************************************************/
static const struct net_device_ops trdev_netdev_ops = {
.ndo_open = tok_open,
.ndo_stop = tok_close,
.ndo_start_xmit = tok_send_packet,
.ndo_set_rx_mode = tok_set_multicast_list,
.ndo_change_mtu = ibmtr_change_mtu,
};
static int __devinit trdev_init(struct net_device *dev)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
SET_PAGE(ti->srb_page);
ti->open_failure = NO ;
dev->netdev_ops = &trdev_netdev_ops;
return 0;
}
/*****************************************************************************/
static int tok_init_card(struct net_device *dev)
{
struct tok_info *ti;
short PIOaddr;
unsigned long i;
PIOaddr = dev->base_addr;
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
ti = netdev_priv(dev);
/* Special processing for first interrupt after reset */
ti->do_tok_int = FIRST_INT;
/* Reset adapter */
writeb(~INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_EVEN);
outb(0, PIOaddr + ADAPTRESET);
schedule_timeout_uninterruptible(TR_RST_TIME); /* wait 50ms */
outb(0, PIOaddr + ADAPTRESETREL);
#ifdef ENABLE_PAGING
if (ti->page_mask)
writeb(SRPR_ENABLE_PAGING,ti->mmio+ACA_OFFSET+ACA_RW+SRPR_EVEN);
#endif
writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN);
i = sleep_on_timeout(&ti->wait_for_reset, 4 * HZ);
return i? 0 : -EAGAIN;
}
/*****************************************************************************/
static int tok_open(struct net_device *dev)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
int i;
/*the case we were left in a failure state during a previous open */
if (ti->open_failure == YES) {
DPRINTK("Last time you were disconnected, how about now?\n");
printk("You can't insert with an ICS connector half-cocked.\n");
}
ti->open_status = CLOSED; /* CLOSED or OPEN */
ti->sap_status = CLOSED; /* CLOSED or OPEN */
ti->open_failure = NO; /* NO or YES */
ti->open_mode = MANUAL; /* MANUAL or AUTOMATIC */
ti->sram_phys &= ~1; /* to reverse what we do in tok_close */
/* init the spinlock */
spin_lock_init(&ti->lock);
init_timer(&ti->tr_timer);
i = tok_init_card(dev);
if (i) return i;
while (1){
tok_open_adapter((unsigned long) dev);
i= interruptible_sleep_on_timeout(&ti->wait_for_reset, 25 * HZ);
/* sig catch: estimate opening adapter takes more than .5 sec*/
if (i>(245*HZ)/10) break; /* fancier than if (i==25*HZ) */
if (i==0) break;
if (ti->open_status == OPEN && ti->sap_status==OPEN) {
netif_start_queue(dev);
DPRINTK("Adapter is up and running\n");
return 0;
}
i=schedule_timeout_interruptible(TR_RETRY_INTERVAL);
/* wait 30 seconds */
if(i!=0) break; /*prob. a signal, like the i>24*HZ case above */
}
outb(0, dev->base_addr + ADAPTRESET);/* kill pending interrupts*/
DPRINTK("TERMINATED via signal\n"); /*BMS useful */
return -EAGAIN;
}
/*****************************************************************************/
#define COMMAND_OFST 0
#define OPEN_OPTIONS_OFST 8
#define NUM_RCV_BUF_OFST 24
#define RCV_BUF_LEN_OFST 26
#define DHB_LENGTH_OFST 28
#define NUM_DHB_OFST 30
#define DLC_MAX_SAP_OFST 32
#define DLC_MAX_STA_OFST 33
static void tok_open_adapter(unsigned long dev_addr)
{
struct net_device *dev = (struct net_device *) dev_addr;
struct tok_info *ti;
int i;
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
ti = netdev_priv(dev);
SET_PAGE(ti->init_srb_page);
writeb(~SRB_RESP_INT, ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_ODD);
for (i = 0; i < sizeof(struct dir_open_adapter); i++)
writeb(0, ti->init_srb + i);
writeb(DIR_OPEN_ADAPTER, ti->init_srb + COMMAND_OFST);
writew(htons(OPEN_PASS_BCON_MAC), ti->init_srb + OPEN_OPTIONS_OFST);
if (ti->ring_speed == 16) {
writew(htons(ti->dhb_size16mb), ti->init_srb + DHB_LENGTH_OFST);
writew(htons(ti->rbuf_cnt16), ti->init_srb + NUM_RCV_BUF_OFST);
writew(htons(ti->rbuf_len16), ti->init_srb + RCV_BUF_LEN_OFST);
} else {
writew(htons(ti->dhb_size4mb), ti->init_srb + DHB_LENGTH_OFST);
writew(htons(ti->rbuf_cnt4), ti->init_srb + NUM_RCV_BUF_OFST);
writew(htons(ti->rbuf_len4), ti->init_srb + RCV_BUF_LEN_OFST);
}
writeb(NUM_DHB, /* always 2 */ ti->init_srb + NUM_DHB_OFST);
writeb(DLC_MAX_SAP, ti->init_srb + DLC_MAX_SAP_OFST);
writeb(DLC_MAX_STA, ti->init_srb + DLC_MAX_STA_OFST);
ti->srb = ti->init_srb; /* We use this one in the interrupt handler */
ti->srb_page = ti->init_srb_page;
DPRINTK("Opening adapter: Xmit bfrs: %d X %d, Rcv bfrs: %d X %d\n",
readb(ti->init_srb + NUM_DHB_OFST),
ntohs(readw(ti->init_srb + DHB_LENGTH_OFST)),
ntohs(readw(ti->init_srb + NUM_RCV_BUF_OFST)),
ntohs(readw(ti->init_srb + RCV_BUF_LEN_OFST)));
writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN);
writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
}
/*****************************************************************************/
static void open_sap(unsigned char type, struct net_device *dev)
{
int i;
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
SET_PAGE(ti->srb_page);
for (i = 0; i < sizeof(struct dlc_open_sap); i++)
writeb(0, ti->srb + i);
#define MAX_I_FIELD_OFST 14
#define SAP_VALUE_OFST 16
#define SAP_OPTIONS_OFST 17
#define STATION_COUNT_OFST 18
writeb(DLC_OPEN_SAP, ti->srb + COMMAND_OFST);
writew(htons(MAX_I_FIELD), ti->srb + MAX_I_FIELD_OFST);
writeb(SAP_OPEN_IND_SAP | SAP_OPEN_PRIORITY, ti->srb+ SAP_OPTIONS_OFST);
writeb(SAP_OPEN_STATION_CNT, ti->srb + STATION_COUNT_OFST);
writeb(type, ti->srb + SAP_VALUE_OFST);
writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
}
/*****************************************************************************/
static void tok_set_multicast_list(struct net_device *dev)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
struct netdev_hw_addr *ha;
unsigned char address[4];
int i;
/*BMS the next line is CRUCIAL or you may be sad when you */
/*BMS ifconfig tr down or hot unplug a PCMCIA card ??hownowbrowncow*/
if (/*BMSHELPdev->start == 0 ||*/ ti->open_status != OPEN) return;
address[0] = address[1] = address[2] = address[3] = 0;
netdev_for_each_mc_addr(ha, dev) {
address[0] |= ha->addr[2];
address[1] |= ha->addr[3];
address[2] |= ha->addr[4];
address[3] |= ha->addr[5];
}
SET_PAGE(ti->srb_page);
for (i = 0; i < sizeof(struct srb_set_funct_addr); i++)
writeb(0, ti->srb + i);
#define FUNCT_ADDRESS_OFST 6
writeb(DIR_SET_FUNC_ADDR, ti->srb + COMMAND_OFST);
for (i = 0; i < 4; i++)
writeb(address[i], ti->srb + FUNCT_ADDRESS_OFST + i);
writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
#if TR_VERBOSE
DPRINTK("Setting functional address: ");
for (i=0;i<4;i++) printk("%02X ", address[i]);
printk("\n");
#endif
}
/*****************************************************************************/
#define STATION_ID_OFST 4
static netdev_tx_t tok_send_packet(struct sk_buff *skb,
struct net_device *dev)
{
struct tok_info *ti;
unsigned long flags;
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
ti = netdev_priv(dev);
netif_stop_queue(dev);
/* lock against other CPUs */
spin_lock_irqsave(&(ti->lock), flags);
/* Save skb; we'll need it when the adapter asks for the data */
ti->current_skb = skb;
SET_PAGE(ti->srb_page);
writeb(XMIT_UI_FRAME, ti->srb + COMMAND_OFST);
writew(ti->exsap_station_id, ti->srb + STATION_ID_OFST);
writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
spin_unlock_irqrestore(&(ti->lock), flags);
return NETDEV_TX_OK;
}
/*****************************************************************************/
static int tok_close(struct net_device *dev)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
/* Important for PCMCIA hot unplug, otherwise, we'll pull the card, */
/* unloading the module from memory, and then if a timer pops, ouch */
del_timer_sync(&ti->tr_timer);
outb(0, dev->base_addr + ADAPTRESET);
ti->sram_phys |= 1;
ti->open_status = CLOSED;
netif_stop_queue(dev);
DPRINTK("Adapter is closed.\n");
return 0;
}
/*****************************************************************************/
#define RETCODE_OFST 2
#define OPEN_ERROR_CODE_OFST 6
#define ASB_ADDRESS_OFST 8
#define SRB_ADDRESS_OFST 10
#define ARB_ADDRESS_OFST 12
#define SSB_ADDRESS_OFST 14
static char *printphase[]= {"Lobe media test","Physical insertion",
"Address verification","Roll call poll","Request Parameters"};
static char *printerror[]={"Function failure","Signal loss","Reserved",
"Frequency error","Timeout","Ring failure","Ring beaconing",
"Duplicate node address",
"Parameter request-retry count exceeded","Remove received",
"IMPL force received","Duplicate modifier",
"No monitor detected","Monitor contention failed for RPL"};
static void __iomem *map_address(struct tok_info *ti, unsigned index, __u8 *page)
{
if (ti->page_mask) {
*page = (index >> 8) & ti->page_mask;
index &= ~(ti->page_mask << 8);
}
return ti->sram_virt + index;
}
static void dir_open_adapter (struct net_device *dev)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
unsigned char ret_code;
__u16 err;
ti->srb = map_address(ti,
ntohs(readw(ti->init_srb + SRB_ADDRESS_OFST)),
&ti->srb_page);
ti->ssb = map_address(ti,
ntohs(readw(ti->init_srb + SSB_ADDRESS_OFST)),
&ti->ssb_page);
ti->arb = map_address(ti,
ntohs(readw(ti->init_srb + ARB_ADDRESS_OFST)),
&ti->arb_page);
ti->asb = map_address(ti,
ntohs(readw(ti->init_srb + ASB_ADDRESS_OFST)),
&ti->asb_page);
ti->current_skb = NULL;
ret_code = readb(ti->init_srb + RETCODE_OFST);
err = ntohs(readw(ti->init_srb + OPEN_ERROR_CODE_OFST));
if (!ret_code) {
ti->open_status = OPEN; /* TR adapter is now available */
if (ti->open_mode == AUTOMATIC) {
DPRINTK("Adapter reopened.\n");
}
writeb(~SRB_RESP_INT, ti->mmio+ACA_OFFSET+ACA_RESET+ISRP_ODD);
open_sap(EXTENDED_SAP, dev);
return;
}
ti->open_failure = YES;
if (ret_code == 7){
if (err == 0x24) {
if (!ti->auto_speedsave) {
DPRINTK("Open failed: Adapter speed must match "
"ring speed if Automatic Ring Speed Save is "
"disabled.\n");
ti->open_action = FAIL;
}else
DPRINTK("Retrying open to adjust to "
"ring speed, ");
} else if (err == 0x2d) {
DPRINTK("Physical Insertion: No Monitor Detected, ");
printk("retrying after %ds delay...\n",
TR_RETRY_INTERVAL/HZ);
} else if (err == 0x11) {
DPRINTK("Lobe Media Function Failure (0x11), ");
printk(" retrying after %ds delay...\n",
TR_RETRY_INTERVAL/HZ);
} else {
char **prphase = printphase;
char **prerror = printerror;
int pnr = err / 16 - 1;
int enr = err % 16 - 1;
DPRINTK("TR Adapter misc open failure, error code = ");
if (pnr < 0 || pnr >= ARRAY_SIZE(printphase) ||
enr < 0 ||
enr >= ARRAY_SIZE(printerror))
printk("0x%x, invalid Phase/Error.", err);
else
printk("0x%x, Phase: %s, Error: %s\n", err,
prphase[pnr], prerror[enr]);
printk(" retrying after %ds delay...\n",
TR_RETRY_INTERVAL/HZ);
}
} else DPRINTK("open failed: ret_code = %02X..., ", ret_code);
if (ti->open_action != FAIL) {
if (ti->open_mode==AUTOMATIC){
ti->open_action = REOPEN;
ibmtr_reset_timer(&(ti->tr_timer), dev);
return;
}
wake_up(&ti->wait_for_reset);
return;
}
DPRINTK("FAILURE, CAPUT\n");
}
/******************************************************************************/
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 15:55:46 +02:00
static irqreturn_t tok_interrupt(int irq, void *dev_id)
{
unsigned char status;
/* unsigned char status_even ; */
struct tok_info *ti;
struct net_device *dev;
#ifdef ENABLE_PAGING
unsigned char save_srpr;
#endif
dev = dev_id;
#if TR_VERBOSE
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 15:55:46 +02:00
DPRINTK("Int from tok_driver, dev : %p irq%d\n", dev,irq);
#endif
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
ti = netdev_priv(dev);
if (ti->sram_phys & 1)
return IRQ_NONE; /* PCMCIA card extraction flag */
spin_lock(&(ti->lock));
#ifdef ENABLE_PAGING
save_srpr = readb(ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN);
#endif
/* Disable interrupts till processing is finished */
writeb((~INT_ENABLE), ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_EVEN);
/* Reset interrupt for ISA boards */
if (ti->adapter_int_enable)
outb(0, ti->adapter_int_enable);
else /* used for PCMCIA cards */
outb(0, ti->global_int_enable);
if (ti->do_tok_int == FIRST_INT){
initial_tok_int(dev);
#ifdef ENABLE_PAGING
writeb(save_srpr, ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN);
#endif
spin_unlock(&(ti->lock));
return IRQ_HANDLED;
}
/* Begin interrupt handler HERE inline to avoid the extra
levels of logic and call depth for the original solution. */
status = readb(ti->mmio + ACA_OFFSET + ACA_RW + ISRP_ODD);
/*BMSstatus_even = readb (ti->mmio + ACA_OFFSET + ACA_RW + ISRP_EVEN) */
/*BMSdebugprintk("tok_interrupt: ISRP_ODD = 0x%x ISRP_EVEN = 0x%x\n", */
/*BMS status,status_even); */
if (status & ADAP_CHK_INT) {
int i;
void __iomem *check_reason;
__u8 check_reason_page = 0;
check_reason = map_address(ti,
ntohs(readw(ti->mmio+ ACA_OFFSET+ACA_RW + WWCR_EVEN)),
&check_reason_page);
SET_PAGE(check_reason_page);
DPRINTK("Adapter check interrupt\n");
DPRINTK("8 reason bytes follow: ");
for (i = 0; i < 8; i++, check_reason++)
printk("%02X ", (int) readb(check_reason));
printk("\n");
writeb(~ADAP_CHK_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD);
status = readb(ti->mmio + ACA_OFFSET + ACA_RW + ISRA_EVEN);
DPRINTK("ISRA_EVEN == 0x02%x\n",status);
ti->open_status = CLOSED;
ti->sap_status = CLOSED;
ti->open_mode = AUTOMATIC;
netif_carrier_off(dev);
netif_stop_queue(dev);
ti->open_action = RESTART;
outb(0, dev->base_addr + ADAPTRESET);
ibmtr_reset_timer(&(ti->tr_timer), dev);/*BMS try to reopen*/
spin_unlock(&(ti->lock));
return IRQ_HANDLED;
}
if (readb(ti->mmio + ACA_OFFSET + ACA_RW + ISRP_EVEN)
& (TCR_INT | ERR_INT | ACCESS_INT)) {
DPRINTK("adapter error: ISRP_EVEN : %02x\n",
(int)readb(ti->mmio+ ACA_OFFSET + ACA_RW + ISRP_EVEN));
writeb(~(TCR_INT | ERR_INT | ACCESS_INT),
ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_EVEN);
status= readb(ti->mmio+ ACA_OFFSET + ACA_RW + ISRA_EVEN);/*BMS*/
DPRINTK("ISRA_EVEN == 0x02%x\n",status);/*BMS*/
writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN);
#ifdef ENABLE_PAGING
writeb(save_srpr, ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN);
#endif
spin_unlock(&(ti->lock));
return IRQ_HANDLED;
}
if (status & SRB_RESP_INT) { /* SRB response */
SET_PAGE(ti->srb_page);
#if TR_VERBOSE
DPRINTK("SRB resp: cmd=%02X rsp=%02X\n",
readb(ti->srb), readb(ti->srb + RETCODE_OFST));
#endif
switch (readb(ti->srb)) { /* SRB command check */
case XMIT_DIR_FRAME:{
unsigned char xmit_ret_code;
xmit_ret_code = readb(ti->srb + RETCODE_OFST);
if (xmit_ret_code == 0xff) break;
DPRINTK("error on xmit_dir_frame request: %02X\n",
xmit_ret_code);
if (ti->current_skb) {
dev_kfree_skb_irq(ti->current_skb);
ti->current_skb = NULL;
}
/*dev->tbusy = 0;*/
netif_wake_queue(dev);
if (ti->readlog_pending)
ibmtr_readlog(dev);
break;
}
case XMIT_UI_FRAME:{
unsigned char xmit_ret_code;
xmit_ret_code = readb(ti->srb + RETCODE_OFST);
if (xmit_ret_code == 0xff) break;
DPRINTK("error on xmit_ui_frame request: %02X\n",
xmit_ret_code);
if (ti->current_skb) {
dev_kfree_skb_irq(ti->current_skb);
ti->current_skb = NULL;
}
netif_wake_queue(dev);
if (ti->readlog_pending)
ibmtr_readlog(dev);
break;
}
case DIR_OPEN_ADAPTER:
dir_open_adapter(dev);
break;
case DLC_OPEN_SAP:
if (readb(ti->srb + RETCODE_OFST)) {
DPRINTK("open_sap failed: ret_code = %02X, "
"retrying\n",
(int) readb(ti->srb + RETCODE_OFST));
ti->open_action = REOPEN;
ibmtr_reset_timer(&(ti->tr_timer), dev);
break;
}
ti->exsap_station_id = readw(ti->srb + STATION_ID_OFST);
ti->sap_status = OPEN;/* TR adapter is now available */
if (ti->open_mode==MANUAL){
wake_up(&ti->wait_for_reset);
break;
}
netif_wake_queue(dev);
netif_carrier_on(dev);
break;
case DIR_INTERRUPT:
case DIR_MOD_OPEN_PARAMS:
case DIR_SET_GRP_ADDR:
case DIR_SET_FUNC_ADDR:
case DLC_CLOSE_SAP:
if (readb(ti->srb + RETCODE_OFST))
DPRINTK("error on %02X: %02X\n",
(int) readb(ti->srb + COMMAND_OFST),
(int) readb(ti->srb + RETCODE_OFST));
break;
case DIR_READ_LOG:
if (readb(ti->srb + RETCODE_OFST)){
DPRINTK("error on dir_read_log: %02X\n",
(int) readb(ti->srb + RETCODE_OFST));
netif_wake_queue(dev);
break;
}
#if IBMTR_DEBUG_MESSAGES
#define LINE_ERRORS_OFST 0
#define INTERNAL_ERRORS_OFST 1
#define BURST_ERRORS_OFST 2
#define AC_ERRORS_OFST 3
#define ABORT_DELIMITERS_OFST 4
#define LOST_FRAMES_OFST 6
#define RECV_CONGEST_COUNT_OFST 7
#define FRAME_COPIED_ERRORS_OFST 8
#define FREQUENCY_ERRORS_OFST 9
#define TOKEN_ERRORS_OFST 10
DPRINTK("Line errors %02X, Internal errors %02X, "
"Burst errors %02X\n" "A/C errors %02X, "
"Abort delimiters %02X, Lost frames %02X\n"
"Receive congestion count %02X, "
"Frame copied errors %02X\nFrequency errors %02X, "
"Token errors %02X\n",
(int) readb(ti->srb + LINE_ERRORS_OFST),
(int) readb(ti->srb + INTERNAL_ERRORS_OFST),
(int) readb(ti->srb + BURST_ERRORS_OFST),
(int) readb(ti->srb + AC_ERRORS_OFST),
(int) readb(ti->srb + ABORT_DELIMITERS_OFST),
(int) readb(ti->srb + LOST_FRAMES_OFST),
(int) readb(ti->srb + RECV_CONGEST_COUNT_OFST),
(int) readb(ti->srb + FRAME_COPIED_ERRORS_OFST),
(int) readb(ti->srb + FREQUENCY_ERRORS_OFST),
(int) readb(ti->srb + TOKEN_ERRORS_OFST));
#endif
netif_wake_queue(dev);
break;
default:
DPRINTK("Unknown command %02X encountered\n",
(int) readb(ti->srb));
} /* end switch SRB command check */
writeb(~SRB_RESP_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD);
} /* if SRB response */
if (status & ASB_FREE_INT) { /* ASB response */
SET_PAGE(ti->asb_page);
#if TR_VERBOSE
DPRINTK("ASB resp: cmd=%02X\n", readb(ti->asb));
#endif
switch (readb(ti->asb)) { /* ASB command check */
case REC_DATA:
case XMIT_UI_FRAME:
case XMIT_DIR_FRAME:
break;
default:
DPRINTK("unknown command in asb %02X\n",
(int) readb(ti->asb));
} /* switch ASB command check */
if (readb(ti->asb + 2) != 0xff) /* checks ret_code */
DPRINTK("ASB error %02X in cmd %02X\n",
(int) readb(ti->asb + 2), (int) readb(ti->asb));
writeb(~ASB_FREE_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD);
} /* if ASB response */
#define STATUS_OFST 6
#define NETW_STATUS_OFST 6
if (status & ARB_CMD_INT) { /* ARB response */
SET_PAGE(ti->arb_page);
#if TR_VERBOSE
DPRINTK("ARB resp: cmd=%02X\n", readb(ti->arb));
#endif
switch (readb(ti->arb)) { /* ARB command check */
case DLC_STATUS:
DPRINTK("DLC_STATUS new status: %02X on station %02X\n",
ntohs(readw(ti->arb + STATUS_OFST)),
ntohs(readw(ti->arb+ STATION_ID_OFST)));
break;
case REC_DATA:
tr_rx(dev);
break;
case RING_STAT_CHANGE:{
unsigned short ring_status;
ring_status= ntohs(readw(ti->arb + NETW_STATUS_OFST));
if (ibmtr_debug_trace & TRC_INIT)
DPRINTK("Ring Status Change...(0x%x)\n",
ring_status);
if(ring_status& (REMOVE_RECV|AUTO_REMOVAL|LOBE_FAULT)){
netif_stop_queue(dev);
netif_carrier_off(dev);
DPRINTK("Remove received, or Auto-removal error"
", or Lobe fault\n");
DPRINTK("We'll try to reopen the closed adapter"
" after a %d second delay.\n",
TR_RETRY_INTERVAL/HZ);
/*I was confused: I saw the TR reopening but */
/*forgot:with an RJ45 in an RJ45/ICS adapter */
/*but adapter not in the ring, the TR will */
/* open, and then soon close and come here. */
ti->open_mode = AUTOMATIC;
ti->open_status = CLOSED; /*12/2000 BMS*/
ti->open_action = REOPEN;
ibmtr_reset_timer(&(ti->tr_timer), dev);
} else if (ring_status & LOG_OVERFLOW) {
if(netif_queue_stopped(dev))
ti->readlog_pending = 1;
else
ibmtr_readlog(dev);
}
break;
}
case XMIT_DATA_REQ:
tr_tx(dev);
break;
default:
DPRINTK("Unknown command %02X in arb\n",
(int) readb(ti->arb));
break;
} /* switch ARB command check */
writeb(~ARB_CMD_INT, ti->mmio+ ACA_OFFSET+ACA_RESET + ISRP_ODD);
writeb(ARB_FREE, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
} /* if ARB response */
if (status & SSB_RESP_INT) { /* SSB response */
unsigned char retcode;
SET_PAGE(ti->ssb_page);
#if TR_VERBOSE
DPRINTK("SSB resp: cmd=%02X rsp=%02X\n",
readb(ti->ssb), readb(ti->ssb + 2));
#endif
switch (readb(ti->ssb)) { /* SSB command check */
case XMIT_DIR_FRAME:
case XMIT_UI_FRAME:
retcode = readb(ti->ssb + 2);
if (retcode && (retcode != 0x22))/* checks ret_code */
DPRINTK("xmit ret_code: %02X xmit error code: "
"%02X\n",
(int)retcode, (int)readb(ti->ssb + 6));
else
dev->stats.tx_packets++;
break;
case XMIT_XID_CMD:
DPRINTK("xmit xid ret_code: %02X\n",
(int) readb(ti->ssb + 2));
default:
DPRINTK("Unknown command %02X in ssb\n",
(int) readb(ti->ssb));
} /* SSB command check */
writeb(~SSB_RESP_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD);
writeb(SSB_FREE, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
} /* if SSB response */
#ifdef ENABLE_PAGING
writeb(save_srpr, ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN);
#endif
writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN);
spin_unlock(&(ti->lock));
return IRQ_HANDLED;
} /*tok_interrupt */
/*****************************************************************************/
#define INIT_STATUS_OFST 1
#define INIT_STATUS_2_OFST 2
#define ENCODED_ADDRESS_OFST 8
static void initial_tok_int(struct net_device *dev)
{
__u32 encoded_addr, hw_encoded_addr;
struct tok_info *ti;
unsigned char init_status; /*BMS 12/2000*/
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
ti = netdev_priv(dev);
ti->do_tok_int = NOT_FIRST;
/* we assign the shared-ram address for ISA devices */
writeb(ti->sram_base, ti->mmio + ACA_OFFSET + ACA_RW + RRR_EVEN);
#ifndef PCMCIA
ti->sram_virt = ioremap(((__u32)ti->sram_base << 12), ti->avail_shared_ram);
#endif
ti->init_srb = map_address(ti,
ntohs(readw(ti->mmio + ACA_OFFSET + WRBR_EVEN)),
&ti->init_srb_page);
if (ti->page_mask && ti->avail_shared_ram == 127) {
void __iomem *last_512;
__u8 last_512_page=0;
int i;
last_512 = map_address(ti, 0xfe00, &last_512_page);
/* initialize high section of ram (if necessary) */
SET_PAGE(last_512_page);
for (i = 0; i < 512; i++)
writeb(0, last_512 + i);
}
SET_PAGE(ti->init_srb_page);
#if TR_VERBOSE
{
int i;
DPRINTK("ti->init_srb_page=0x%x\n", ti->init_srb_page);
DPRINTK("init_srb(%p):", ti->init_srb );
for (i = 0; i < 20; i++)
printk("%02X ", (int) readb(ti->init_srb + i));
printk("\n");
}
#endif
hw_encoded_addr = readw(ti->init_srb + ENCODED_ADDRESS_OFST);
encoded_addr = ntohs(hw_encoded_addr);
init_status= /*BMS 12/2000 check for shallow mode possibility (Turbo)*/
readb(ti->init_srb+offsetof(struct srb_init_response,init_status));
/*printk("Initial interrupt: init_status= 0x%02x\n",init_status);*/
ti->ring_speed = init_status & 0x01 ? 16 : 4;
DPRINTK("Initial interrupt : %d Mbps, shared RAM base %08x.\n",
ti->ring_speed, (unsigned int)dev->mem_start);
ti->auto_speedsave = (readb(ti->init_srb+INIT_STATUS_2_OFST) & 4) != 0;
if (ti->open_mode == MANUAL) wake_up(&ti->wait_for_reset);
else tok_open_adapter((unsigned long)dev);
} /*initial_tok_int() */
/*****************************************************************************/
#define CMD_CORRELATE_OFST 1
#define DHB_ADDRESS_OFST 6
#define FRAME_LENGTH_OFST 6
#define HEADER_LENGTH_OFST 8
#define RSAP_VALUE_OFST 9
static void tr_tx(struct net_device *dev)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
struct trh_hdr *trhdr = (struct trh_hdr *) ti->current_skb->data;
unsigned int hdr_len;
__u32 dhb=0,dhb_base;
void __iomem *dhbuf = NULL;
unsigned char xmit_command;
int i,dhb_len=0x4000,src_len,src_offset;
struct trllc *llc;
struct srb_xmit xsrb;
__u8 dhb_page = 0;
__u8 llc_ssap;
SET_PAGE(ti->asb_page);
if (readb(ti->asb+RETCODE_OFST) != 0xFF) DPRINTK("ASB not free !!!\n");
/* in providing the transmit interrupts, is telling us it is ready for
data and providing a shared memory address for us to stuff with data.
Here we compute the effective address where we will place data.
*/
SET_PAGE(ti->arb_page);
dhb=dhb_base=ntohs(readw(ti->arb + DHB_ADDRESS_OFST));
if (ti->page_mask) {
dhb_page = (dhb_base >> 8) & ti->page_mask;
dhb=dhb_base & ~(ti->page_mask << 8);
}
dhbuf = ti->sram_virt + dhb;
/* Figure out the size of the 802.5 header */
if (!(trhdr->saddr[0] & 0x80)) /* RIF present? */
hdr_len = sizeof(struct trh_hdr) - TR_MAXRIFLEN;
else
hdr_len = ((ntohs(trhdr->rcf) & TR_RCF_LEN_MASK) >> 8)
+ sizeof(struct trh_hdr) - TR_MAXRIFLEN;
llc = (struct trllc *) (ti->current_skb->data + hdr_len);
llc_ssap = llc->ssap;
SET_PAGE(ti->srb_page);
memcpy_fromio(&xsrb, ti->srb, sizeof(xsrb));
SET_PAGE(ti->asb_page);
xmit_command = xsrb.command;
writeb(xmit_command, ti->asb + COMMAND_OFST);
writew(xsrb.station_id, ti->asb + STATION_ID_OFST);
writeb(llc_ssap, ti->asb + RSAP_VALUE_OFST);
writeb(xsrb.cmd_corr, ti->asb + CMD_CORRELATE_OFST);
writeb(0, ti->asb + RETCODE_OFST);
if ((xmit_command == XMIT_XID_CMD) || (xmit_command == XMIT_TEST_CMD)) {
writew(htons(0x11), ti->asb + FRAME_LENGTH_OFST);
writeb(0x0e, ti->asb + HEADER_LENGTH_OFST);
SET_PAGE(dhb_page);
writeb(AC, dhbuf);
writeb(LLC_FRAME, dhbuf + 1);
for (i = 0; i < TR_ALEN; i++)
writeb((int) 0x0FF, dhbuf + i + 2);
for (i = 0; i < TR_ALEN; i++)
writeb(0, dhbuf + i + TR_ALEN + 2);
writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
return;
}
/*
* the token ring packet is copied from sk_buff to the adapter
* buffer identified in the command data received with the interrupt.
*/
writeb(hdr_len, ti->asb + HEADER_LENGTH_OFST);
writew(htons(ti->current_skb->len), ti->asb + FRAME_LENGTH_OFST);
src_len=ti->current_skb->len;
src_offset=0;
dhb=dhb_base;
while(1) {
if (ti->page_mask) {
dhb_page=(dhb >> 8) & ti->page_mask;
dhb=dhb & ~(ti->page_mask << 8);
dhb_len=0x4000-dhb; /* remaining size of this page */
}
dhbuf = ti->sram_virt + dhb;
SET_PAGE(dhb_page);
if (src_len > dhb_len) {
memcpy_toio(dhbuf,&ti->current_skb->data[src_offset],
dhb_len);
src_len -= dhb_len;
src_offset += dhb_len;
dhb_base+=dhb_len;
dhb=dhb_base;
continue;
}
memcpy_toio(dhbuf, &ti->current_skb->data[src_offset], src_len);
break;
}
writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
dev->stats.tx_bytes += ti->current_skb->len;
dev_kfree_skb_irq(ti->current_skb);
ti->current_skb = NULL;
netif_wake_queue(dev);
if (ti->readlog_pending)
ibmtr_readlog(dev);
} /*tr_tx */
/*****************************************************************************/
#define RECEIVE_BUFFER_OFST 6
#define LAN_HDR_LENGTH_OFST 8
#define DLC_HDR_LENGTH_OFST 9
#define DSAP_OFST 0
#define SSAP_OFST 1
#define LLC_OFST 2
#define PROTID_OFST 3
#define ETHERTYPE_OFST 6
static void tr_rx(struct net_device *dev)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
__u32 rbuffer;
void __iomem *rbuf, *rbufdata, *llc;
__u8 rbuffer_page = 0;
unsigned char *data;
unsigned int rbuffer_len, lan_hdr_len, hdr_len, ip_len, length;
unsigned char dlc_hdr_len;
struct sk_buff *skb;
unsigned int skb_size = 0;
int IPv4_p = 0;
unsigned int chksum = 0;
struct iphdr *iph;
struct arb_rec_req rarb;
SET_PAGE(ti->arb_page);
memcpy_fromio(&rarb, ti->arb, sizeof(rarb));
rbuffer = ntohs(rarb.rec_buf_addr) ;
rbuf = map_address(ti, rbuffer, &rbuffer_page);
SET_PAGE(ti->asb_page);
if (readb(ti->asb + RETCODE_OFST) !=0xFF) DPRINTK("ASB not free !!!\n");
writeb(REC_DATA, ti->asb + COMMAND_OFST);
writew(rarb.station_id, ti->asb + STATION_ID_OFST);
writew(rarb.rec_buf_addr, ti->asb + RECEIVE_BUFFER_OFST);
lan_hdr_len = rarb.lan_hdr_len;
if (lan_hdr_len > sizeof(struct trh_hdr)) {
DPRINTK("Linux cannot handle greater than 18 bytes RIF\n");
return;
} /*BMS I added this above just to be very safe */
dlc_hdr_len = readb(ti->arb + DLC_HDR_LENGTH_OFST);
hdr_len = lan_hdr_len + sizeof(struct trllc) + sizeof(struct iphdr);
SET_PAGE(rbuffer_page);
llc = rbuf + offsetof(struct rec_buf, data) + lan_hdr_len;
#if TR_VERBOSE
DPRINTK("offsetof data: %02X lan_hdr_len: %02X\n",
(__u32) offsetof(struct rec_buf, data), (unsigned int) lan_hdr_len);
DPRINTK("llc: %08X rec_buf_addr: %04X dev->mem_start: %lX\n",
llc, ntohs(rarb.rec_buf_addr), dev->mem_start);
DPRINTK("dsap: %02X, ssap: %02X, llc: %02X, protid: %02X%02X%02X, "
"ethertype: %04X\n",
(int) readb(llc + DSAP_OFST), (int) readb(llc + SSAP_OFST),
(int) readb(llc + LLC_OFST), (int) readb(llc + PROTID_OFST),
(int) readb(llc+PROTID_OFST+1),(int)readb(llc+PROTID_OFST + 2),
(int) ntohs(readw(llc + ETHERTYPE_OFST)));
#endif
if (readb(llc + offsetof(struct trllc, llc)) != UI_CMD) {
SET_PAGE(ti->asb_page);
writeb(DATA_LOST, ti->asb + RETCODE_OFST);
dev->stats.rx_dropped++;
writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
return;
}
length = ntohs(rarb.frame_len);
if (readb(llc + DSAP_OFST) == EXTENDED_SAP &&
readb(llc + SSAP_OFST) == EXTENDED_SAP &&
length >= hdr_len) IPv4_p = 1;
#if TR_VERBOSE
#define SADDR_OFST 8
#define DADDR_OFST 2
if (!IPv4_p) {
void __iomem *trhhdr = rbuf + offsetof(struct rec_buf, data);
u8 saddr[6];
u8 daddr[6];
int i;
for (i = 0 ; i < 6 ; i++)
saddr[i] = readb(trhhdr + SADDR_OFST + i);
for (i = 0 ; i < 6 ; i++)
daddr[i] = readb(trhhdr + DADDR_OFST + i);
DPRINTK("Probably non-IP frame received.\n");
DPRINTK("ssap: %02X dsap: %02X "
"saddr: %pM daddr: %pM\n",
readb(llc + SSAP_OFST), readb(llc + DSAP_OFST),
saddr, daddr);
}
#endif
/*BMS handle the case she comes in with few hops but leaves with many */
skb_size=length-lan_hdr_len+sizeof(struct trh_hdr)+sizeof(struct trllc);
if (!(skb = dev_alloc_skb(skb_size))) {
DPRINTK("out of memory. frame dropped.\n");
dev->stats.rx_dropped++;
SET_PAGE(ti->asb_page);
writeb(DATA_LOST, ti->asb + offsetof(struct asb_rec, ret_code));
writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
return;
}
/*BMS again, if she comes in with few but leaves with many */
skb_reserve(skb, sizeof(struct trh_hdr) - lan_hdr_len);
skb_put(skb, length);
data = skb->data;
rbuffer_len = ntohs(readw(rbuf + offsetof(struct rec_buf, buf_len)));
rbufdata = rbuf + offsetof(struct rec_buf, data);
if (IPv4_p) {
/* Copy the headers without checksumming */
memcpy_fromio(data, rbufdata, hdr_len);
/* Watch for padded packets and bogons */
iph= (struct iphdr *)(data+ lan_hdr_len + sizeof(struct trllc));
ip_len = ntohs(iph->tot_len) - sizeof(struct iphdr);
length -= hdr_len;
if ((ip_len <= length) && (ip_len > 7))
length = ip_len;
data += hdr_len;
rbuffer_len -= hdr_len;
rbufdata += hdr_len;
}
/* Copy the payload... */
#define BUFFER_POINTER_OFST 2
#define BUFFER_LENGTH_OFST 6
for (;;) {
if (ibmtr_debug_trace&TRC_INITV && length < rbuffer_len)
DPRINTK("CURIOUS, length=%d < rbuffer_len=%d\n",
length,rbuffer_len);
if (IPv4_p)
chksum=csum_partial_copy_nocheck((void*)rbufdata,
data,length<rbuffer_len?length:rbuffer_len,chksum);
else
memcpy_fromio(data, rbufdata, rbuffer_len);
rbuffer = ntohs(readw(rbuf+BUFFER_POINTER_OFST)) ;
if (!rbuffer)
break;
rbuffer -= 2;
length -= rbuffer_len;
data += rbuffer_len;
rbuf = map_address(ti, rbuffer, &rbuffer_page);
SET_PAGE(rbuffer_page);
rbuffer_len = ntohs(readw(rbuf + BUFFER_LENGTH_OFST));
rbufdata = rbuf + offsetof(struct rec_buf, data);
}
SET_PAGE(ti->asb_page);
writeb(0, ti->asb + offsetof(struct asb_rec, ret_code));
writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
dev->stats.rx_bytes += skb->len;
dev->stats.rx_packets++;
skb->protocol = tr_type_trans(skb, dev);
if (IPv4_p) {
skb->csum = chksum;
skb->ip_summed = CHECKSUM_COMPLETE;
}
netif_rx(skb);
} /*tr_rx */
/*****************************************************************************/
static void ibmtr_reset_timer(struct timer_list *tmr, struct net_device *dev)
{
tmr->expires = jiffies + TR_RETRY_INTERVAL;
tmr->data = (unsigned long) dev;
tmr->function = tok_rerun;
init_timer(tmr);
add_timer(tmr);
}
/*****************************************************************************/
static void tok_rerun(unsigned long dev_addr)
{
struct net_device *dev = (struct net_device *)dev_addr;
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
if ( ti->open_action == RESTART){
ti->do_tok_int = FIRST_INT;
outb(0, dev->base_addr + ADAPTRESETREL);
#ifdef ENABLE_PAGING
if (ti->page_mask)
writeb(SRPR_ENABLE_PAGING,
ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN);
#endif
writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN);
} else
tok_open_adapter(dev_addr);
}
/*****************************************************************************/
static void ibmtr_readlog(struct net_device *dev)
{
struct tok_info *ti;
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
ti = netdev_priv(dev);
ti->readlog_pending = 0;
SET_PAGE(ti->srb_page);
writeb(DIR_READ_LOG, ti->srb);
writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN);
writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD);
netif_stop_queue(dev);
}
/*****************************************************************************/
static int ibmtr_change_mtu(struct net_device *dev, int mtu)
{
dev->priv to netdev_priv(dev), drivers/net/tokenring/ Replacing accesses to dev->priv to netdev_priv(dev). The replacment is safe when netdev_priv is used to access a private structure that is right next to the net_device structure in memory. Cf http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/de19321bcd94dbb8/0d74a4adcd6177bd This is the case when the net_device structure was allocated with a call to alloc_netdev or one of its derivative. Here is an excerpt of the semantic patch that performs the transformation @ rule1 @ type T; struct net_device *dev; @@ dev = ( alloc_netdev | alloc_etherdev | alloc_trdev ) (sizeof(T), ...) @ rule1bis @ struct net_device *dev; expression E; @@ dev->priv = E @ rule2 depends on rule1 && !rule1bis @ struct net_device *dev; type rule1.T; @@ - (T*) dev->priv + netdev_priv(dev) PS: I have performed the same transformation on the whole kernel and it affects around 70 files, most of them in drivers/net/. Should I split my patch for each subnet directories ? (wireless/, wan/, etc) Thanks to Thomas Surrel for helping me refining my semantic patch. Signed-off-by: Yoann Padioleau <padator@wanadoo.fr> 3c359.c | 58 +++++++++++++++++++++++++++++----------------------------- ibmtr.c | 38 +++++++++++++++++++------------------- lanstreamer.c | 32 ++++++++++++++++---------------- madgemc.c | 4 ++-- olympic.c | 36 ++++++++++++++++++------------------ tmspci.c | 4 ++-- 6 files changed, 86 insertions(+), 86 deletions(-) Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-07-23 15:18:21 +02:00
struct tok_info *ti = netdev_priv(dev);
if (ti->ring_speed == 16 && mtu > ti->maxmtu16)
return -EINVAL;
if (ti->ring_speed == 4 && mtu > ti->maxmtu4)
return -EINVAL;
dev->mtu = mtu;
return 0;
}
/*****************************************************************************/
#ifdef MODULE
/* 3COM 3C619C supports 8 interrupts, 32 I/O ports */
static struct net_device *dev_ibmtr[IBMTR_MAX_ADAPTERS];
static int io[IBMTR_MAX_ADAPTERS] = { 0xa20, 0xa24 };
static int irq[IBMTR_MAX_ADAPTERS];
static int mem[IBMTR_MAX_ADAPTERS];
MODULE_LICENSE("GPL");
module_param_array(io, int, NULL, 0);
module_param_array(irq, int, NULL, 0);
module_param_array(mem, int, NULL, 0);
static int __init ibmtr_init(void)
{
int i;
int count=0;
find_turbo_adapters(io);
for (i = 0; i < IBMTR_MAX_ADAPTERS && io[i]; i++) {
struct net_device *dev;
irq[i] = 0;
mem[i] = 0;
dev = alloc_trdev(sizeof(struct tok_info));
if (dev == NULL) {
if (i == 0)
return -ENOMEM;
break;
}
dev->base_addr = io[i];
dev->irq = irq[i];
dev->mem_start = mem[i];
if (ibmtr_probe_card(dev)) {
free_netdev(dev);
continue;
}
dev_ibmtr[i] = dev;
count++;
}
if (count) return 0;
printk("ibmtr: register_netdev() returned non-zero.\n");
return -EIO;
}
module_init(ibmtr_init);
static void __exit ibmtr_cleanup(void)
{
int i;
for (i = 0; i < IBMTR_MAX_ADAPTERS; i++){
if (!dev_ibmtr[i])
continue;
unregister_netdev(dev_ibmtr[i]);
ibmtr_cleanup_card(dev_ibmtr[i]);
free_netdev(dev_ibmtr[i]);
}
}
module_exit(ibmtr_cleanup);
#endif