patch-2.1.88 linux/drivers/acorn/block/mfmhd.c
Next file: linux/drivers/acorn/net/Config.in
Previous file: linux/drivers/acorn/block/ide-rapide.c
Back to the patch index
Back to the overall index
- Lines: 1550
- Date:
Mon Jan 26 14:03:39 1998
- Orig file:
v2.1.87/linux/drivers/acorn/block/mfmhd.c
- Orig date:
Wed Dec 31 16:00:00 1969
diff -u --recursive --new-file v2.1.87/linux/drivers/acorn/block/mfmhd.c linux/drivers/acorn/block/mfmhd.c
@@ -0,0 +1,1549 @@
+/*
+ * linux/arch/arm/drivers/block/mfmhd.c
+ *
+ * Copyright (C) 1995, 1996 Russell King, Dave Alan Gilbert (gilbertd@cs.man.ac.uk)
+ *
+ * MFM hard drive code [experimental]
+ */
+
+/*
+ * Change list:
+ *
+ * 3/2/96:DAG: Started a change list :-)
+ * Set the hardsect_size pointers up since we are running 256 byte
+ * sectors
+ * Added DMA code, put it into the rw_intr
+ * Moved RCAL out of generic interrupt code - don't want to do it
+ * while DMA'ing - its now in individual handlers.
+ * Took interrupt handlers off task queue lists and called
+ * directly - not sure of implications.
+ *
+ * 18/2/96:DAG: Well its reading OK I think, well enough for image file code
+ * to find the image file; but now I've discovered that I actually
+ * have to put some code in for image files.
+ *
+ * Added stuff for image files; seems to work, but I've not
+ * got a multisegment image file (I don't think!).
+ * Put in a hack (yep a real hack) for multiple cylinder reads.
+ * Not convinced its working.
+ *
+ * 5/4/96:DAG: Added asm/hardware.h and use IOC_ macros
+ * Rewrote dma code in mfm.S (again!) - now takes a word at a time
+ * from main RAM for speed; still doesn't feel speedy!
+ *
+ * 20/4/96:DAG: After rewriting mfm.S a heck of a lot of times and speeding
+ * things up, I've finally figured out why its so damn slow.
+ * Linux is only reading a block at a time, and so you never
+ * get more than 1K per disc revoloution ~=60K/second.
+ *
+ * 27/4/96:DAG: On Russell's advice I change ll_rw_blk.c to ask it to
+ * join adjacent blocks together. Everything falls flat on its
+ * face.
+ * Four hours of debugging later; I hadn't realised that
+ * ll_rw_blk would be so generous as to join blocks whose
+ * results aren't going into consecutive buffers.
+ *
+ * OK; severe rehacking of mfm_rw_interrupt; now end_request's
+ * as soon as its DMA'd each request. Odd thing is that
+ * we are sometimes getting interrupts where we are not transferring
+ * any data; why? Is that what happens when you miss? I doubt
+ * it; are we too fast? No - its just at command ends. Got 240K/s
+ * better than before, but RiscOS hits 480K/s
+ *
+ * 25/6/96:RMK: Fixed init code to allow the MFM podule to work. Increased the
+ * number of errors for my Miniscribe drive (8425).
+ *
+ * 30/6/96:DAG: Russell suggested that a check drive 0 might turn the LEDs off
+ * - so in request_done just before it clears Busy it sends a
+ * check drive 0 - and the LEDs go off!!!!
+ *
+ * Added test for mainboard controller. - Removes need for separate
+ * define.
+ *
+ * 13/7/96:DAG: Changed hardware sectore size to 512 in attempt to make
+ * IM drivers work.
+ * 21/7/96:DAG: Took out old image file stuff (accessing it now produces an IO
+ * error.)
+ *
+ * 17/8/96:DAG: Ran through indent -kr -i8; evil - all my nice 2 character indents
+ * gone :-( Hand modified afterwards.
+ * Took out last remains of the older image map system.
+ *
+ * 22/9/96:DAG: Changed mfm.S so it will carry on DMA'ing til; BSY is dropped
+ * Changed mfm_rw_intr so that it doesn't follow the error
+ * code until BSY is dropped. Nope - still broke. Problem
+ * may revolve around when it reads the results for the error
+ * number?
+ *
+ *16/11/96:DAG: Modified for 2.0.18; request_irq changed
+ *
+ *17/12/96:RMK: Various cleanups, reorganisation, and the changes for new IO system.
+ * Improved probe for onboard MFM chip - it was hanging on my A5k.
+ * Added autodetect CHS code such that we don't rely on the presence
+ * of an ADFS boot block. Added ioport resource manager calls so
+ * that we don't clash with already-running hardware (eg. RiscPC Ether
+ * card slots if someone tries this)!
+ *
+ * 17/1/97:RMK: Upgraded to 2.1 kernels.
+ */
+
+/*
+ * Possible enhancements:
+ * Multi-thread the code so that it is possible that while one drive
+ * is seeking, the other one can be reading data/seeking as well.
+ * This would be a performance boost with dual drive systems.
+ */
+
+#include <linux/module.h>
+#include <linux/config.h>
+#include <linux/sched.h>
+#include <linux/fs.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/timer.h>
+#include <linux/tqueue.h>
+#include <linux/mm.h>
+#include <linux/errno.h>
+#include <linux/genhd.h>
+#include <linux/major.h>
+#include <linux/ioport.h>
+
+#include <asm/system.h>
+#include <asm/io.h>
+#include <asm/irq-no.h>
+#include <asm/uaccess.h>
+#include <asm/delay.h>
+#include <asm/dma.h>
+#include <asm/hardware.h>
+#include <asm/ecard.h>
+
+#define MFM_DISK_MAJOR 13
+#undef XT_DISK_MAJOR
+#define XT_DISK_MAJOR -1
+#define MAJOR_NR MFM_DISK_MAJOR
+#include "blk.h"
+
+/*
+ * This sort of stuff should be in a header file shared with ide.c, hd.c, xd.c etc
+ */
+#ifndef HDIO_GETGEO
+#define HDIO_GETGEO 0x301
+struct hd_geometry {
+ unsigned char heads;
+ unsigned char sectors;
+ unsigned short cylinders;
+ unsigned long start;
+};
+#endif
+
+
+/*
+ * Configuration section
+ *
+ * This is the maximum number of drives that we accept
+ */
+#define MFM_MAXDRIVES 2
+/*
+ * Linux I/O address of onboard MFM controller or 0 to disable this
+ */
+#define ONBOARD_MFM_ADDRESS ((0x002d0000 >> 2) | 0x80000000)
+/*
+ * Uncomment this to enable debugging in the MFM driver...
+ */
+#ifndef DEBUG
+/*#define DEBUG */
+#endif
+/*
+ * List of card types that we recognise
+ */
+static const card_ids mfm_cids[] = {
+ { MANU_ACORN, PROD_ACORN_MFM },
+ { 0xffff, 0xffff }
+};
+/*
+ * End of configuration
+ */
+
+
+/*
+ * This structure contains all information to do with a particular physical
+ * device.
+ */
+struct mfm_info {
+ unsigned char sectors;
+ unsigned char heads;
+ unsigned short cylinders;
+ unsigned short lowcurrent;
+ unsigned short precomp;
+#define NO_TRACK -1
+#define NEED_1_RECAL -2
+#define NEED_2_RECAL -3
+ int cylinder;
+ unsigned int access_count;
+ unsigned int busy;
+ struct {
+ char recal;
+ char report;
+ char abort;
+ } errors;
+} mfm_info[MFM_MAXDRIVES];
+
+#define MFM_DRV_INFO mfm_info[raw_cmd.dev]
+
+static struct hd_struct mfm[MFM_MAXDRIVES << 6];
+static int mfm_sizes[MFM_MAXDRIVES << 6];
+static int mfm_blocksizes[MFM_MAXDRIVES << 6];
+static int mfm_sectsizes[MFM_MAXDRIVES << 6];
+static struct wait_queue *mfm_wait_open = NULL;
+
+/* Stuff from the assembly routines */
+extern unsigned int hdc63463_baseaddress; /* Controller base address */
+extern unsigned int hdc63463_irqpolladdress; /* Address to read to test for int */
+extern unsigned int hdc63463_irqpollmask; /* Mask for irq register */
+extern unsigned int hdc63463_dataptr; /* Pointer to kernel data space to DMA */
+extern int hdc63463_dataleft; /* Number of bytes left to transfer */
+
+
+
+
+static int lastspecifieddrive;
+static unsigned Busy;
+
+static unsigned int PartFragRead; /* The number of sectors which have been read
+ during a partial read split over two
+ cylinders. If 0 it means a partial
+ read did not occur. */
+
+static unsigned int PartFragRead_RestartBlock; /* Where to restart on a split access */
+static unsigned int PartFragRead_SectorsLeft; /* Where to restart on a split access */
+
+static int Sectors256LeftInCurrent; /* i.e. 256 byte sectors left in current */
+static int SectorsLeftInRequest; /* i.e. blocks left in the thing mfm_request was called for */
+static int Copy_Sector; /* The 256 byte sector we are currently at - fragments need to know
+ where to take over */
+static char *Copy_buffer;
+
+
+static void mfm_seek(void);
+static void mfm_rerequest(void);
+static void mfm_request(void);
+static int mfm_reread_partitions(kdev_t dev);
+static void mfm_specify (void);
+static void issue_request(int dev, unsigned int block, unsigned int nsect,
+ struct request *req);
+
+#define mfm_init xd_init
+#define mfm_setup xd_setup
+
+static unsigned int mfm_addr; /* Controller address */
+static unsigned int mfm_IRQPollLoc; /* Address to read for IRQ information */
+static unsigned int mfm_irqenable; /* Podule IRQ enable location */
+static unsigned char mfm_irq; /* Interrupt number */
+static int mfm_drives = 0; /* drives available */
+static int mfm_status = 0; /* interrupt status */
+static int *errors;
+
+static struct rawcmd {
+ unsigned int dev;
+ unsigned int cylinder;
+ unsigned int head;
+ unsigned int sector;
+ unsigned int cmdtype;
+ unsigned int cmdcode;
+ unsigned char cmddata[16];
+ unsigned int cmdlen;
+} raw_cmd;
+
+static unsigned char result[16];
+
+static struct cont {
+ void (*interrupt) (void); /* interrupt handler */
+ void (*error) (void); /* error handler */
+ void (*redo) (void); /* redo handler */
+ void (*done) (int st); /* done handler */
+} *cont = NULL;
+
+static struct tq_struct mfm_tq = {0, 0, (void (*)(void *)) NULL, 0};
+
+int number_mfm_drives = 1;
+
+/* ------------------------------------------------------------------------------------------ */
+/*
+ * From the HD63463 data sheet from Hitachi Ltd.
+ */
+
+#define MFM_COMMAND (mfm_addr + 0)
+#define MFM_DATAOUT (mfm_addr + 1)
+#define MFM_STATUS (mfm_addr + 8)
+#define MFM_DATAIN (mfm_addr + 9)
+
+#define CMD_ABT 0xF0 /* Abort */
+#define CMD_SPC 0xE8 /* Specify */
+#define CMD_TST 0xE0 /* Test */
+#define CMD_RCLB 0xC8 /* Recalibrate */
+#define CMD_SEK 0xC0 /* Seek */
+#define CMD_WFS 0xAB /* Write Format Skew */
+#define CMD_WFM 0xA3 /* Write Format */
+#define CMD_MTB 0x90 /* Memory to buffer */
+#define CMD_CMPD 0x88 /* Compare data */
+#define CMD_WD 0x87 /* Write data */
+#define CMD_RED 0x70 /* Read erroneous data */
+#define CMD_RIS 0x68 /* Read ID skew */
+#define CMD_FID 0x61 /* Find ID */
+#define CMD_RID 0x60 /* Read ID */
+#define CMD_BTM 0x50 /* Buffer to memory */
+#define CMD_CKD 0x48 /* Check data */
+#define CMD_RD 0x40 /* Read data */
+#define CMD_OPBW 0x38 /* Open buffer write */
+#define CMD_OPBR 0x30 /* Open buffer read */
+#define CMD_CKV 0x28 /* Check drive */
+#define CMD_CKE 0x20 /* Check ECC */
+#define CMD_POD 0x18 /* Polling disable */
+#define CMD_POL 0x10 /* Polling enable */
+#define CMD_RCAL 0x08 /* Recall */
+
+#define STAT_BSY 0x8000 /* Busy */
+#define STAT_CPR 0x4000 /* Command Parameter Rejection */
+#define STAT_CED 0x2000 /* Command end */
+#define STAT_SED 0x1000 /* Seek end */
+#define STAT_DER 0x0800 /* Drive error */
+#define STAT_ABN 0x0400 /* Abnormal end */
+#define STAT_POL 0x0200 /* Polling */
+
+/* ------------------------------------------------------------------------------------------ */
+#ifdef DEBUG
+static void console_printf(const char *fmt,...)
+{
+ static char buffer[2048]; /* Arbitary! */
+ extern void console_print(const char *);
+ unsigned long flags;
+ va_list ap;
+
+ save_flags_cli(flags);
+
+ va_start(ap, fmt);
+ vsprintf(buffer, fmt, ap);
+ console_print(buffer);
+ va_end(fmt);
+
+ restore_flags(flags);
+}; /* console_printf */
+
+#define DBG(x...) console_printf(x)
+#else
+#define DBG(x...)
+#endif
+
+static void print_status(void)
+{
+ char *error;
+ static char *errors[] = {
+ "no error",
+ "command aborted",
+ "invalid command",
+ "parameter error",
+ "not initialised",
+ "rejected TEST",
+ "no useld",
+ "write fault",
+ "not ready",
+ "no scp",
+ "in seek",
+ "invalid NCA",
+ "invalid step rate",
+ "seek error",
+ "over run",
+ "invalid PHA",
+ "data field EEC error",
+ "data field CRC error",
+ "error corrected",
+ "data field fatal error",
+ "no data am",
+ "not hit",
+ "ID field CRC error",
+ "time over",
+ "no ID am",
+ "not writable"
+ };
+ if (result[1] < 0x65)
+ error = errors[result[1] >> 2];
+ else
+ error = "unknown";
+ printk("(");
+ if (mfm_status & STAT_BSY) printk("BSY ");
+ if (mfm_status & STAT_CPR) printk("CPR ");
+ if (mfm_status & STAT_CED) printk("CED ");
+ if (mfm_status & STAT_SED) printk("SED ");
+ if (mfm_status & STAT_DER) printk("DER ");
+ if (mfm_status & STAT_ABN) printk("ABN ");
+ if (mfm_status & STAT_POL) printk("POL ");
+ printk(") SSB = %X (%s)\n", result[1], error);
+
+}
+
+/* ------------------------------------------------------------------------------------- */
+
+static void issue_command(int command, unsigned char *cmdb, int len)
+{
+ int status;
+#ifdef DEBUG
+ int i;
+ console_printf("issue_command: %02X: ", command);
+ for (i = 0; i < len; i++)
+ console_printf("%02X ", cmdb[i]);
+ console_printf("\n");
+#endif
+
+ do {
+ status = inw(MFM_STATUS);
+ } while (status & (STAT_BSY | STAT_POL));
+ DBG("issue_command: status after pol/bsy loop: %02X:\n ", status >> 8);
+
+ if (status & (STAT_CPR | STAT_CED | STAT_SED | STAT_DER | STAT_ABN)) {
+ outw(CMD_RCAL, MFM_COMMAND);
+ while (inw(MFM_STATUS) & STAT_BSY);
+ }
+ status = inw(MFM_STATUS);
+ DBG("issue_command: status before parameter issue: %02X:\n ", status >> 8);
+
+ while (len > 0) {
+ outw(cmdb[1] | (cmdb[0] << 8), MFM_DATAOUT);
+ len -= 2;
+ cmdb += 2;
+ }
+ status = inw(MFM_STATUS);
+ DBG("issue_command: status before command issue: %02X:\n ", status >> 8);
+
+ outw(command, MFM_COMMAND);
+ status = inw(MFM_STATUS);
+ DBG("issue_command: status immediatly after command issue: %02X:\n ", status >> 8);
+}
+
+static void wait_for_completion(void)
+{
+ while ((mfm_status = inw(MFM_STATUS)) & STAT_BSY);
+}
+
+static void wait_for_command_end(void)
+{
+ int i;
+
+ while (!((mfm_status = inw(MFM_STATUS)) & STAT_CED));
+
+ for (i = 0; i < 16;) {
+ int in;
+ in = inw(MFM_DATAIN);
+ result[i++] = in >> 8;
+ result[i++] = in;
+ }
+ outw (CMD_RCAL, MFM_COMMAND);
+}
+
+/* ------------------------------------------------------------------------------------- */
+
+static void mfm_rw_intr(void)
+{
+ int old_status; /* Holds status on entry, we read to see if the command just finished */
+#ifdef DEBUG
+ console_printf("mfm_rw_intr...dataleft=%d\n", hdc63463_dataleft);
+ print_status();
+#endif
+
+ /* Now don't handle the error until BSY drops */
+ if ((mfm_status & (STAT_DER | STAT_ABN)) && ((mfm_status&STAT_BSY)==0)) {
+ /* Something has gone wrong - lets try that again */
+ outw(CMD_RCAL, MFM_COMMAND); /* Clear interrupt condition */
+ if (cont) {
+ DBG("mfm_rw_intr: DER/ABN err\n");
+ cont->error();
+ cont->redo();
+ };
+ return;
+ };
+
+ /* OK so what ever happend its not an error, now I reckon we are left between
+ a choice of command end or some data which is ready to be collected */
+ /* I think we have to transfer data while the interrupt line is on and its
+ not any other type of interrupt */
+ if (CURRENT->cmd == WRITE) {
+ extern void hdc63463_writedma(void);
+ if ((hdc63463_dataleft <= 0) && (!(mfm_status & STAT_CED))) {
+ printk("mfm_rw_intr: Apparent DMA write request when no more to DMA\n");
+ if (cont) {
+ cont->error();
+ cont->redo();
+ };
+ return;
+ };
+ hdc63463_writedma();
+ } else {
+ extern void hdc63463_readdma(void);
+ if ((hdc63463_dataleft <= 0) && (!(mfm_status & STAT_CED))) {
+ printk("mfm_rw_intr: Apparent DMA read request when no more to DMA\n");
+ if (cont) {
+ cont->error();
+ cont->redo();
+ };
+ return;
+ };
+ DBG("Going to try read dma..............status=0x%x, buffer=%p\n", mfm_status, hdc63463_dataptr);
+ hdc63463_readdma();
+ }; /* Read */
+
+ if (hdc63463_dataptr != ((unsigned int) Copy_buffer + 256)) {
+ /* If we didn't actually manage to get any data on this interrupt - but why? We got the interrupt */
+ /* Ah - well looking at the status its just when we get command end; so no problem */
+ /*console_printf("mfm: dataptr mismatch. dataptr=0x%08x Copy_buffer+256=0x%08p\n",
+ hdc63463_dataptr,Copy_buffer+256);
+ print_status(); */
+ } else {
+ Sectors256LeftInCurrent--;
+ Copy_buffer += 256;
+ Copy_Sector++;
+
+ /* We have come to the end of this request */
+ if (!Sectors256LeftInCurrent) {
+ DBG("mfm: end_request for CURRENT=0x%p CURRENT(sector=%d current_nr_sectors=%d nr_sectors=%d)\n",
+ CURRENT, CURRENT->sector, CURRENT->current_nr_sectors, CURRENT->nr_sectors);
+
+ CURRENT->nr_sectors -= CURRENT->current_nr_sectors;
+ CURRENT->sector += CURRENT->current_nr_sectors;
+ SectorsLeftInRequest -= CURRENT->current_nr_sectors;
+
+ end_request(1);
+ if (SectorsLeftInRequest) {
+ hdc63463_dataptr = (unsigned int) CURRENT->buffer;
+ Copy_buffer = CURRENT->buffer;
+ Sectors256LeftInCurrent = CURRENT->current_nr_sectors * 2;
+ errors = &(CURRENT->errors);
+ /* These should match the present calculations of the next logical sector
+ on the device
+ Copy_Sector=CURRENT->sector*2; */
+
+ if (Copy_Sector != CURRENT->sector * 2)
+#ifdef DEBUG
+ /*console_printf*/printk("mfm: Copy_Sector mismatch. Copy_Sector=%d CURRENT->sector*2=%d\n",
+ Copy_Sector, CURRENT->sector * 2);
+#else
+ printk("mfm: Copy_Sector mismatch! Eek!\n");
+#endif
+ }; /* CURRENT */
+ }; /* Sectors256LeftInCurrent */
+ };
+
+ old_status = mfm_status;
+ mfm_status = inw(MFM_STATUS);
+ if (mfm_status & (STAT_DER | STAT_ABN)) {
+ /* Something has gone wrong - lets try that again */
+ if (cont) {
+ DBG("mfm_rw_intr: DER/ABN error\n");
+ cont->error();
+ cont->redo();
+ };
+ return;
+ };
+
+ /* If this code wasn't entered due to command_end but there is
+ now a command end we must read the command results out. If it was
+ entered like this then mfm_interrupt_handler would have done the
+ job. */
+ if ((!((old_status & (STAT_CPR | STAT_BSY)) == STAT_CPR)) &&
+ ((mfm_status & (STAT_CPR | STAT_BSY)) == STAT_CPR)) {
+ int len = 0;
+ while (len < 16) {
+ int in;
+ in = inw(MFM_DATAIN);
+ result[len++] = in >> 8;
+ result[len++] = in;
+ };
+ }; /* Result read */
+
+ /*console_printf ("mfm_rw_intr nearexit [%02X]\n", inb(mfm_IRQPollLoc)); */
+
+ /* If end of command move on */
+ if (mfm_status & (STAT_CED)) {
+ outw(CMD_RCAL, MFM_COMMAND); /* Clear interrupt condition */
+ /* End of command - trigger the next command */
+ if (cont) {
+ cont->done(1);
+ }
+ DBG("mfm_rw_intr: returned from cont->done\n");
+ } else {
+ /* Its going to generate another interrupt */
+ SET_INTR(mfm_rw_intr);
+ };
+}
+
+static void mfm_setup_rw(void)
+{
+ DBG("setting up for rw...\n");
+
+ SET_INTR(mfm_rw_intr);
+ issue_command(raw_cmd.cmdcode, raw_cmd.cmddata, raw_cmd.cmdlen);
+}
+
+static void mfm_recal_intr(void)
+{
+#ifdef DEBUG
+ console_printf("recal intr - status = ");
+ print_status();
+#endif
+ outw(CMD_RCAL, MFM_COMMAND); /* Clear interrupt condition */
+ if (mfm_status & (STAT_DER | STAT_ABN)) {
+ printk("recal failed\n");
+ MFM_DRV_INFO.cylinder = NEED_2_RECAL;
+ if (cont) {
+ cont->error();
+ cont->redo();
+ }
+ return;
+ }
+ /* Thats seek end - we are finished */
+ if (mfm_status & STAT_SED) {
+ issue_command(CMD_POD, NULL, 0);
+ MFM_DRV_INFO.cylinder = 0;
+ mfm_seek();
+ return;
+ }
+ /* Command end without seek end (see data sheet p.20) for parallel seek
+ - we have to send a POL command to wait for the seek */
+ if (mfm_status & STAT_CED) {
+ SET_INTR(mfm_recal_intr);
+ issue_command(CMD_POL, NULL, 0);
+ return;
+ }
+ printk("recal: unknown status\n");
+}
+
+static void mfm_seek_intr(void)
+{
+#ifdef DEBUG
+ console_printf("seek intr - status = ");
+ print_status();
+#endif
+ outw(CMD_RCAL, MFM_COMMAND); /* Clear interrupt condition */
+ if (mfm_status & (STAT_DER | STAT_ABN)) {
+ printk("seek failed\n");
+ MFM_DRV_INFO.cylinder = NEED_2_RECAL;
+ if (cont) {
+ cont->error();
+ cont->redo();
+ }
+ return;
+ }
+ if (mfm_status & STAT_SED) {
+ issue_command(CMD_POD, NULL, 0);
+ MFM_DRV_INFO.cylinder = raw_cmd.cylinder;
+ mfm_seek();
+ return;
+ }
+ if (mfm_status & STAT_CED) {
+ SET_INTR(mfm_seek_intr);
+ issue_command(CMD_POL, NULL, 0);
+ return;
+ }
+ printk("seek: unknown status\n");
+}
+
+/* IDEA2 seems to work better - its what RiscOS sets my
+ * disc to - on its SECOND call to specify!
+ */
+#define IDEA2
+#ifndef IDEA2
+#define SPEC_SL 0x16
+#define SPEC_SH 0xa9 /* Step pulse high=21, Record Length=001 (256 bytes) */
+#else
+#define SPEC_SL 0x00 /* OM2 - SL - step pulse low */
+#define SPEC_SH 0x21 /* Step pulse high=4, Record Length=001 (256 bytes) */
+#endif
+
+static void mfm_setupspecify (int drive, unsigned char *cmdb)
+{
+ cmdb[0] = 0x1f; /* OM0 - !SECT,!MOD,!DIF,PADP,ECD,CRCP,CRCI,ACOR */
+ cmdb[1] = 0xc3; /* OM1 - DTM,BRST,!CEDM,!SEDM,!DERM,0,AMEX,PSK */
+ cmdb[2] = SPEC_SL; /* OM2 - SL - step pulse low */
+ cmdb[3] = (number_mfm_drives == 1) ? 0x02 : 0x06; /* 1 or 2 drives */
+ cmdb[4] = 0xfc | ((mfm_info[drive].cylinders - 1) >> 8);/* RW time over/high part of number of cylinders */
+ cmdb[5] = mfm_info[drive].cylinders - 1; /* low part of number of cylinders */
+ cmdb[6] = mfm_info[drive].heads - 1; /* Number of heads */
+ cmdb[7] = mfm_info[drive].sectors - 1; /* Number of sectors */
+ cmdb[8] = SPEC_SH;
+ cmdb[9] = 0x0a; /* gap length 1 */
+ cmdb[10] = 0x0d; /* gap length 2 */
+ cmdb[11] = 0x0c; /* gap length 3 */
+ cmdb[12] = (mfm_info[drive].precomp - 1) >> 8; /* pre comp cylinder */
+ cmdb[13] = mfm_info[drive].precomp - 1;
+ cmdb[14] = (mfm_info[drive].lowcurrent - 1) >> 8; /* Low current cylinder */
+ cmdb[15] = mfm_info[drive].lowcurrent - 1;
+}
+
+static void mfm_specify (void)
+{
+ unsigned char cmdb[16];
+
+ DBG("specify...dev=%d lastspecified=%d\n", raw_cmd.dev, lastspecifieddrive);
+ mfm_setupspecify (raw_cmd.dev, cmdb);
+
+ issue_command (CMD_SPC, cmdb, 16);
+ /* Ensure that we will do another specify if we move to the other drive */
+ lastspecifieddrive = raw_cmd.dev;
+ wait_for_completion();
+}
+
+static void mfm_seek(void)
+{
+ unsigned char cmdb[4];
+
+ DBG("seeking...\n");
+ if (MFM_DRV_INFO.cylinder < 0) {
+ SET_INTR(mfm_recal_intr);
+ DBG("mfm_seek: about to call specify\n");
+ mfm_specify (); /* DAG added this */
+
+ cmdb[0] = raw_cmd.dev + 1;
+ cmdb[1] = 0;
+
+ issue_command(CMD_RCLB, cmdb, 2);
+ return;
+ }
+ if (MFM_DRV_INFO.cylinder != raw_cmd.cylinder) {
+ cmdb[0] = raw_cmd.dev + 1;
+ cmdb[1] = 0; /* raw_cmd.head; DAG: My data sheet says this should be 0 */
+ cmdb[2] = raw_cmd.cylinder >> 8;
+ cmdb[3] = raw_cmd.cylinder;
+
+ SET_INTR(mfm_seek_intr);
+ issue_command(CMD_SEK, cmdb, 4);
+ } else
+ mfm_setup_rw();
+}
+
+static void mfm_initialise(void)
+{
+ DBG("init...\n");
+ mfm_seek();
+}
+
+static void request_done(int uptodate)
+{
+ DBG("mfm:request_done\n");
+ if (uptodate) {
+ unsigned char block[2] = {0, 0};
+
+ /* Apparently worked - lets check bytes left to DMA */
+ if (hdc63463_dataleft != (PartFragRead_SectorsLeft * 256)) {
+ printk("mfm: request_done - dataleft=%d - should be %d - Eek!\n", hdc63463_dataleft, PartFragRead_SectorsLeft * 256);
+ end_request(0);
+ Busy = 0;
+ };
+ /* Potentially this means that we've done; but we might be doing
+ a partial access, (over two cylinders) or we may have a number
+ of fragments in an image file. First lets deal with partial accesss
+ */
+ if (PartFragRead) {
+ /* Yep - a partial access */
+
+ /* and issue the remainder */
+ issue_request(MINOR(CURRENT->rq_dev), PartFragRead_RestartBlock, PartFragRead_SectorsLeft, CURRENT);
+ return;
+ }
+
+ /* ah well - perhaps there is another fragment to go */
+
+ /* Increment pointers/counts to start of next fragment */
+ if (SectorsLeftInRequest > 0) printk("mfm: SectorsLeftInRequest>0 - Eek! Shouldn't happen!\n");
+
+ /* No - its the end of the line */
+ /* end_request's should have happened at the end of sector DMAs */
+ /* Turns Drive LEDs off - may slow it down? */
+ if (!CURRENT)
+ issue_command(CMD_CKV, block, 2);
+
+ Busy = 0;
+ DBG("request_done: About to mfm_request\n");
+ /* Next one please */
+ mfm_request(); /* Moved from mfm_rw_intr */
+ DBG("request_done: returned from mfm_request\n");
+ } else {
+ printk("mfm:request_done: update=0\n");
+ end_request(0);
+ Busy = 0;
+ }
+}
+
+static void error_handler(void)
+{
+ printk("error detected... status = ");
+ print_status();
+ (*errors)++;
+ if (*errors > MFM_DRV_INFO.errors.abort)
+ cont->done(0);
+ if (*errors > MFM_DRV_INFO.errors.recal)
+ MFM_DRV_INFO.cylinder = NEED_2_RECAL;
+}
+
+static void rw_interrupt(void)
+{
+ printk("rw_interrupt\n");
+}
+
+static struct cont rw_cont =
+{
+ rw_interrupt,
+ error_handler,
+ mfm_rerequest,
+ request_done
+};
+
+/*
+ * Actually gets round to issuing the request - note everything at this
+ * point is in 256 byte sectors not Linux 512 byte blocks
+ */
+static void issue_request(int dev, unsigned int block, unsigned int nsect,
+ struct request *req)
+{
+ int track, start_head, start_sector;
+ int sectors_to_next_cyl;
+
+ dev >>= 6;
+
+ track = block / mfm_info[dev].sectors;
+ start_sector = block % mfm_info[dev].sectors;
+ start_head = track % mfm_info[dev].heads;
+
+ /* First get the number of whole tracks which are free before the next
+ track */
+ sectors_to_next_cyl = (mfm_info[dev].heads - (start_head + 1)) * mfm_info[dev].sectors;
+ /* Then add in the number of sectors left on this track */
+ sectors_to_next_cyl += (mfm_info[dev].sectors - start_sector);
+
+ DBG("issue_request: mfm_info[dev].sectors=%d track=%d\n", mfm_info[dev].sectors, track);
+
+ raw_cmd.dev = dev;
+ raw_cmd.sector = start_sector;
+ raw_cmd.head = start_head;
+ raw_cmd.cylinder = track / mfm_info[dev].heads;
+ raw_cmd.cmdtype = CURRENT->cmd;
+ raw_cmd.cmdcode = CURRENT->cmd == WRITE ? CMD_WD : CMD_RD;
+ raw_cmd.cmddata[0] = dev + 1; /* DAG: +1 to get US */
+ raw_cmd.cmddata[1] = raw_cmd.head;
+ raw_cmd.cmddata[2] = raw_cmd.cylinder >> 8;
+ raw_cmd.cmddata[3] = raw_cmd.cylinder;
+ raw_cmd.cmddata[4] = raw_cmd.head;
+ raw_cmd.cmddata[5] = raw_cmd.sector;
+
+ /* Was == and worked - how the heck??? */
+ if (lastspecifieddrive != raw_cmd.dev)
+ mfm_specify ();
+
+ if (nsect <= sectors_to_next_cyl) {
+ raw_cmd.cmddata[6] = nsect >> 8;
+ raw_cmd.cmddata[7] = nsect;
+ PartFragRead = 0; /* All in one */
+ PartFragRead_SectorsLeft = 0; /* Must set this - used in DMA calcs */
+ } else {
+ raw_cmd.cmddata[6] = sectors_to_next_cyl >> 8;
+ raw_cmd.cmddata[7] = sectors_to_next_cyl;
+ PartFragRead = sectors_to_next_cyl; /* only do this many this time */
+ PartFragRead_RestartBlock = block + sectors_to_next_cyl; /* Where to restart from */
+ PartFragRead_SectorsLeft = nsect - sectors_to_next_cyl;
+ }
+ raw_cmd.cmdlen = 8;
+
+ /* Setup DMA pointers */
+ hdc63463_dataptr = (unsigned int) Copy_buffer;
+ hdc63463_dataleft = nsect * 256; /* Better way? */
+
+ DBG("mfm%c: %sing: CHS=%d/%d/%d, sectors=%d, buffer=0x%08lx (%p)\n",
+ raw_cmd.dev + 'a', (CURRENT->cmd == READ) ? "read" : "writ",
+ raw_cmd.cylinder,
+ raw_cmd.head,
+ raw_cmd.sector, nsect, (unsigned long) Copy_buffer, CURRENT);
+
+ cont = &rw_cont;
+ errors = &(CURRENT->errors);
+#if 0
+ mfm_tq.routine = (void (*)(void *)) mfm_initialise;
+ queue_task(&mfm_tq, &tq_immediate);
+ mark_bh(IMMEDIATE_BH);
+#else
+ mfm_initialise();
+#endif
+} /* issue_request */
+
+/*
+ * Called when an error has just happened - need to trick mfm_request
+ * into thinking we weren't busy
+ *
+ * Turn off ints - mfm_request expects them this way
+ */
+static void mfm_rerequest(void)
+{
+ DBG("mfm_rerequest\n");
+ cli();
+ Busy = 0;
+ mfm_request();
+}
+
+static void mfm_request(void)
+{
+ DBG("mfm_request CURRENT=%p Busy=%d\n", CURRENT, Busy);
+
+ if (!CURRENT) {
+ DBG("mfm_request: Exited due to NULL Current 1\n");
+ return;
+ }
+
+ if (CURRENT->rq_status == RQ_INACTIVE) {
+ /* Hmm - seems to be happening a lot on 1.3.45 */
+ /*console_printf("mfm_request: Exited due to INACTIVE Current\n"); */
+ return;
+ }
+
+ /* If we are still processing then return; we will get called again */
+ if (Busy) {
+ /* Again seems to be common in 1.3.45 */
+ /*DBG*/printk("mfm_request: Exiting due to busy\n");
+ return;
+ }
+ Busy = 1;
+
+ while (1) {
+ unsigned int dev, block, nsect;
+
+ DBG("mfm_request: loop start\n");
+ sti();
+
+ DBG("mfm_request: before INIT_REQUEST\n");
+
+ if (!CURRENT) {
+ printk("mfm_request: Exiting due to !CURRENT (pre)\n");
+ CLEAR_INTR;
+ Busy = 0;
+ return;
+ };
+
+ INIT_REQUEST;
+
+ DBG("mfm_request: before arg extraction\n");
+
+ dev = MINOR(CURRENT->rq_dev);
+ block = CURRENT->sector;
+ nsect = CURRENT->nr_sectors;
+#ifdef DEBUG
+ /*if ((dev>>6)==1) */ console_printf("mfm_request: raw vals: dev=%d (block=512 bytes) block=%d nblocks=%d\n", dev, block, nsect);
+#endif
+ if (dev >= (mfm_drives << 6) ||
+ block >= mfm[dev].nr_sects || ((block+nsect) > mfm[dev].nr_sects)) {
+ if (dev >= (mfm_drives << 6))
+ printk("mfm: bad minor number: device=%s\n", kdevname(CURRENT->rq_dev));
+ else
+ printk("mfm%c: bad access: block=%d, count=%d, nr_sects=%ld\n", (dev >> 6)+'a',
+ block, nsect, mfm[dev].nr_sects);
+ printk("mfm: continue 1\n");
+ end_request(0);
+ Busy = 0;
+ continue;
+ }
+
+ block += mfm[dev].start_sect;
+
+ /* DAG: Linux doesn't cope with this - even though it has an array telling
+ it the hardware block size - silly */
+ block <<= 1; /* Now in 256 byte sectors */
+ nsect <<= 1; /* Ditto */
+
+ SectorsLeftInRequest = nsect >> 1;
+ Sectors256LeftInCurrent = CURRENT->current_nr_sectors * 2;
+ Copy_buffer = CURRENT->buffer;
+ Copy_Sector = CURRENT->sector << 1;
+
+ DBG("mfm_request: block after offset=%d\n", block);
+
+ if (CURRENT->cmd != READ && CURRENT->cmd != WRITE) {
+ printk("unknown mfm-command %d\n", CURRENT->cmd);
+ end_request(0);
+ Busy = 0;
+ printk("mfm: continue 4\n");
+ continue;
+ }
+ issue_request(dev, block, nsect, CURRENT);
+
+ break;
+ }
+ DBG("mfm_request: Dropping out bottom\n");
+}
+
+static void do_mfm_request(void)
+{
+ DBG("do_mfm_request: about to mfm_request\n");
+ mfm_request();
+}
+
+static void mfm_interrupt_handler(int unused, void *dev_id, struct pt_regs *regs)
+{
+ void (*handler) (void) = DEVICE_INTR;
+
+ CLEAR_INTR;
+
+ DBG("mfm_interrupt_handler (handler=0x%p)\n", handler);
+
+ mfm_status = inw(MFM_STATUS);
+
+ /* If CPR (Command Parameter Reject) and not busy it means that the command
+ has some return message to give us */
+ if ((mfm_status & (STAT_CPR | STAT_BSY)) == STAT_CPR) {
+ int len = 0;
+ while (len < 16) {
+ int in;
+ in = inw(MFM_DATAIN);
+ result[len++] = in >> 8;
+ result[len++] = in;
+ }
+ }
+ if (handler) {
+ handler();
+ return;
+ }
+ outw (CMD_RCAL, MFM_COMMAND); /* Clear interrupt condition */
+ printk ("mfm: unexpected interrupt - status = ");
+ print_status ();
+ while (1);
+}
+
+
+
+
+
+/*
+ * Tell the user about the drive if we decided it exists. Also,
+ * set the size of the drive.
+ */
+static void mfm_geometry (int drive)
+{
+ if (mfm_info[drive].cylinders)
+ printk ("mfm%c: %dMB CHS=%d/%d/%d LCC=%d RECOMP=%d\n", 'a' + drive,
+ mfm_info[drive].cylinders * mfm_info[drive].heads * mfm_info[drive].sectors / 4096,
+ mfm_info[drive].cylinders, mfm_info[drive].heads, mfm_info[drive].sectors,
+ mfm_info[drive].lowcurrent, mfm_info[drive].precomp);
+ mfm[drive << 6].start_sect = 0;
+ mfm[drive << 6].nr_sects = mfm_info[drive].cylinders * mfm_info[drive].heads * mfm_info[drive].sectors / 2;
+}
+
+#ifdef CONFIG_BLK_DEV_MFM_AUTODETECT
+/*
+ * Attempt to detect a drive and find its geometry. The drive has already been
+ * specified...
+ *
+ * We first recalibrate the disk, then try to probe sectors, heads and then
+ * cylinders. NOTE! the cylinder probe may break drives. The xd disk driver
+ * does something along these lines, so I assume that most drives are up to
+ * this mistreatment...
+ */
+static int mfm_detectdrive (int drive)
+{
+ unsigned int mingeo[3], maxgeo[3];
+ unsigned int attribute, need_recal = 1;
+ unsigned char cmdb[8];
+
+ memset (mingeo, 0, sizeof (mingeo));
+ maxgeo[0] = mfm_info[drive].sectors;
+ maxgeo[1] = mfm_info[drive].heads;
+ maxgeo[2] = mfm_info[drive].cylinders;
+
+ cmdb[0] = drive + 1;
+ cmdb[6] = 0;
+ cmdb[7] = 1;
+ for (attribute = 0; attribute < 3; attribute++) {
+ while (mingeo[attribute] != maxgeo[attribute]) {
+ unsigned int variable;
+
+ variable = (maxgeo[attribute] + mingeo[attribute]) >> 1;
+ cmdb[1] = cmdb[2] = cmdb[3] = cmdb[4] = cmdb[5] = 0;
+
+ if (need_recal) {
+ int tries = 5;
+
+ do {
+ issue_command (CMD_RCLB, cmdb, 2);
+ wait_for_completion ();
+ wait_for_command_end ();
+ if (result[1] == 0x20)
+ break;
+ } while (result[1] && --tries);
+ if (result[1]) {
+ outw (CMD_RCAL, MFM_COMMAND);
+ return 0;
+ }
+ need_recal = 0;
+ }
+
+ switch (attribute) {
+ case 0:
+ cmdb[5] = variable;
+ issue_command (CMD_CMPD, cmdb, 8);
+ break;
+ case 1:
+ cmdb[1] = variable;
+ cmdb[4] = variable;
+ issue_command (CMD_CMPD, cmdb, 8);
+ break;
+ case 2:
+ cmdb[2] = variable >> 8;
+ cmdb[3] = variable;
+ issue_command (CMD_SEK, cmdb, 4);
+ break;
+ }
+ wait_for_completion ();
+ wait_for_command_end ();
+
+ switch (result[1]) {
+ case 0x00:
+ case 0x50:
+ mingeo[attribute] = variable + 1;
+ break;
+
+ case 0x20:
+ outw (CMD_RCAL, MFM_COMMAND);
+ return 0;
+
+ case 0x24:
+ need_recal = 1;
+ default:
+ maxgeo[attribute] = variable;
+ break;
+ }
+ }
+ }
+ mfm_info[drive].cylinders = mingeo[2];
+ mfm_info[drive].lowcurrent = mingeo[2];
+ mfm_info[drive].precomp = mingeo[2] / 2;
+ mfm_info[drive].heads = mingeo[1];
+ mfm_info[drive].sectors = mingeo[0];
+ outw (CMD_RCAL, MFM_COMMAND);
+ return 1;
+}
+#endif
+
+/*
+ * Initialise all drive information for this controller.
+ */
+static int mfm_initdrives(void)
+{
+ int drive;
+
+ if (number_mfm_drives > MFM_MAXDRIVES) {
+ number_mfm_drives = MFM_MAXDRIVES;
+ printk("No. of ADFS MFM drives is greater than MFM_MAXDRIVES - you can't have that many!\n");
+ }
+
+ for (drive = 0; drive < number_mfm_drives; drive++) {
+ mfm_info[drive].lowcurrent = 1;
+ mfm_info[drive].precomp = 1;
+ mfm_info[drive].cylinder = -1;
+ mfm_info[drive].errors.recal = 0;
+ mfm_info[drive].errors.report = 0;
+ mfm_info[drive].errors.abort = 4;
+
+#ifdef CONFIG_BLK_DEV_MFM_AUTODETECT
+ mfm_info[drive].cylinders = 1024;
+ mfm_info[drive].heads = 8;
+ mfm_info[drive].sectors = 64;
+ {
+ unsigned char cmdb[16];
+
+ mfm_setupspecify (drive, cmdb);
+ cmdb[1] &= ~0x81;
+ issue_command (CMD_SPC, cmdb, 16);
+ wait_for_completion ();
+ if (!mfm_detectdrive (drive)) {
+ mfm_info[drive].cylinders = 0;
+ mfm_info[drive].heads = 0;
+ mfm_info[drive].sectors = 0;
+ }
+ cmdb[0] = cmdb[1] = 0;
+ issue_command (CMD_CKV, cmdb, 2);
+ }
+#else
+ mfm_info[drive].cylinders = 1; /* its going to have to figure it out from the partition info */
+ mfm_info[drive].heads = 4;
+ mfm_info[drive].sectors = 32;
+#endif
+ }
+ return number_mfm_drives;
+}
+
+
+
+/*
+ * The 'front' end of the mfm driver follows...
+ */
+
+static int mfm_ioctl(struct inode *inode, struct file *file, u_int cmd, u_long arg)
+{
+ struct hd_geometry *geo = (struct hd_geometry *) arg;
+ kdev_t dev;
+ int device, major, minor, err;
+
+ if (!inode || !(dev = inode->i_rdev))
+ return -EINVAL;
+
+ major = MAJOR(dev);
+ minor = MINOR(dev);
+
+ device = DEVICE_NR(MINOR(inode->i_rdev)), err;
+ if (device >= mfm_drives)
+ return -EINVAL;
+
+ switch (cmd) {
+ case HDIO_GETGEO:
+ if (!arg)
+ return -EINVAL;
+ if (put_user (mfm_info[device].heads, &geo->heads))
+ return -EFAULT;
+ if (put_user (mfm_info[device].sectors, &geo->sectors))
+ return -EFAULT;
+ if (put_user (mfm_info[device].cylinders, &geo->cylinders))
+ return -EFAULT;
+ if (put_user (mfm[minor].start_sect, &geo->start))
+ return -EFAULT;
+ return 0;
+
+ case BLKFLSBUF:
+ if (!suser())
+ return -EACCES;
+ fsync_dev(dev);
+ invalidate_buffers(dev);
+ return 0;
+
+ case BLKRASET:
+ if (!suser())
+ return -EACCES;
+ if (arg > 0xff)
+ return -EINVAL;
+ read_ahead[major] = arg;
+ return 0;
+
+ case BLKRAGET:
+ return put_user(read_ahead[major], (long *)arg);
+
+ case BLKGETSIZE:
+ return put_user (mfm[minor].nr_sects, (long *)arg);
+
+ case BLKFRASET:
+ if (!suser())
+ return -EACCES;
+ max_readahead[major][minor] = arg;
+ return 0;
+
+ case BLKFRAGET:
+ return put_user(max_readahead[major][minor], (long *) arg);
+
+ case BLKSECTGET:
+ return put_user(max_sectors[major][minor], (long *) arg);
+
+ case BLKRRPART:
+ if (!suser())
+ return -EACCES;
+ return mfm_reread_partitions(dev);
+
+ RO_IOCTLS(dev, arg);
+
+ default:
+ return -EINVAL;
+ }
+}
+
+static int mfm_open(struct inode *inode, struct file *file)
+{
+ int dev = DEVICE_NR(MINOR(inode->i_rdev));
+
+ if (dev >= mfm_drives)
+ return -ENODEV;
+
+ MOD_INC_USE_COUNT;
+ while (mfm_info[dev].busy)
+ sleep_on (&mfm_wait_open);
+
+ mfm_info[dev].access_count++;
+ return 0;
+}
+
+/*
+ * Releasing a block device means we sync() it, so that it can safely
+ * be forgotten about...
+ */
+static int mfm_release(struct inode *inode, struct file *file)
+{
+ fsync_dev(inode->i_rdev);
+ mfm_info[DEVICE_NR(MINOR(inode->i_rdev))].access_count--;
+ MOD_DEC_USE_COUNT;
+ return 0;
+}
+
+/*
+ * This is to handle various kernel command line parameters
+ * specific to this driver.
+ */
+void mfm_setup(char *str, int *ints)
+{
+ return;
+}
+
+/*
+ * Set the CHS from the ADFS boot block if it is present. This is not ideal
+ * since if there are any non-ADFS partitions on the disk, this won't work!
+ * Hence, I want to get rid of this...
+ */
+void xd_set_geometry(kdev_t dev, unsigned char secsptrack, unsigned char heads,
+ unsigned long discsize, unsigned int secsize)
+{
+ int drive = MINOR(dev) >> 6;
+
+ if (mfm_info[drive].cylinders == 1) {
+ mfm_info[drive].sectors = secsptrack;
+ mfm_info[drive].heads = heads;
+ mfm_info[drive].cylinders = discsize / (secsptrack * heads * secsize);
+
+ if ((heads < 1) || (mfm_info[drive].cylinders > 1024)) {
+ printk("mfm%c: Insane disc shape! Setting to 512/4/32\n",'a' + (dev >> 6));
+
+ /* These values are fairly arbitary, but are there so that if your
+ * lucky you can pick apart your disc to find out what is going on -
+ * I reckon these figures won't hurt MOST drives
+ */
+ mfm_info[drive].sectors = 32;
+ mfm_info[drive].heads = 4;
+ mfm_info[drive].cylinders = 512;
+ }
+ if (raw_cmd.dev == drive)
+ mfm_specify ();
+ mfm_geometry (drive);
+ }
+}
+
+static void mfm_geninit (struct gendisk *gdev);
+
+static struct gendisk mfm_gendisk = {
+ MAJOR_NR, /* Major number */
+ "mfm", /* Major name */
+ 6, /* Bits to shift to get real from partition */
+ 1 << 6, /* Number of partitions per real */
+ MFM_MAXDRIVES, /* maximum number of real */
+ mfm_geninit, /* init function */
+ mfm, /* hd struct */
+ mfm_sizes, /* block sizes */
+ 0, /* number */
+ (void *) mfm_info, /* internal */
+ NULL /* next */
+};
+
+static void mfm_geninit (struct gendisk *gdev)
+{
+ int i;
+
+ mfm_drives = mfm_initdrives();
+
+ printk("mfm: detected %d hard drive%s\n", mfm_drives, mfm_drives == 1 ? "" : "s");
+ gdev->nr_real = mfm_drives;
+
+ for (i = 0; i < mfm_drives; i++)
+ mfm_geometry (i);
+
+ if (request_irq(mfm_irq, mfm_interrupt_handler, SA_INTERRUPT, "MFM harddisk", NULL))
+ printk("mfm: unable to get IRQ%d\n", mfm_irq);
+
+ if (mfm_irqenable)
+ outw(0x80, mfm_irqenable); /* Required to enable IRQs from MFM podule */
+
+ for (i = 0; i < (MFM_MAXDRIVES << 6); i++) {
+ mfm_blocksizes[i] = 1024; /* Can't increase this - if you do all hell breaks loose */
+ mfm_sectsizes[i] = 512;
+ }
+ blksize_size[MAJOR_NR] = mfm_blocksizes;
+ hardsect_size[MAJOR_NR] = mfm_sectsizes;
+}
+
+static struct file_operations mfm_fops =
+{
+ NULL, /* lseek - default */
+ block_read, /* read - general block-dev read */
+ block_write, /* write - general block-dev write */
+ NULL, /* readdir - bad */
+ NULL, /* poll */
+ mfm_ioctl, /* ioctl */
+ NULL, /* mmap */
+ mfm_open, /* open */
+ mfm_release, /* release */
+ block_fsync, /* fsync */
+ NULL, /* fasync */
+ NULL, /* check_media_change */
+ NULL /* revalidate */
+};
+
+
+static struct expansion_card *ecs;
+
+/*
+ * See if there is a controller at the address presently at mfm_addr
+ *
+ * We check to see if the controller is busy - if it is, we abort it first,
+ * and check that the chip is no longer busy after at least 180 clock cycles.
+ * We then issue a command and check that the BSY or CPR bits are set.
+ */
+static int mfm_probecontroller (unsigned int mfm_addr)
+{
+ if (check_region (mfm_addr, 10))
+ return 0;
+
+ if (inw (MFM_STATUS) & STAT_BSY) {
+ outw (CMD_ABT, MFM_COMMAND);
+ udelay (50);
+ if (inw (MFM_STATUS) & STAT_BSY)
+ return 0;
+ }
+
+ if (inw (MFM_STATUS) & STAT_CED)
+ outw (CMD_RCAL, MFM_COMMAND);
+
+ outw (CMD_SEK, MFM_COMMAND);
+
+ if (inw (MFM_STATUS) & (STAT_BSY | STAT_CPR)) {
+ unsigned int count = 2000;
+ while (inw (MFM_STATUS) & STAT_BSY) {
+ udelay (500);
+ if (!--count)
+ return 0;
+ }
+
+ outw (CMD_RCAL, MFM_COMMAND);
+ }
+ return 1;
+}
+
+/*
+ * Look for a MFM controller - first check the motherboard, then the podules
+ * The podules have an extra interrupt enable that needs to be played with
+ *
+ * The HDC is accessed at MEDIUM IOC speeds.
+ */
+int mfm_init (void)
+{
+ unsigned char irqmask;
+
+ if (register_blkdev(MAJOR_NR, "mfm", &mfm_fops)) {
+ printk("mfm_init: unable to get major number %d\n", MAJOR_NR);
+ return -1;
+ }
+
+ if (mfm_probecontroller(ONBOARD_MFM_ADDRESS)) {
+ mfm_addr = ONBOARD_MFM_ADDRESS;
+ mfm_IRQPollLoc = IOC_IRQSTATB;
+ mfm_irqenable = 0;
+ mfm_irq = IRQ_HARDDISK;
+ irqmask = 0x08; /* IL3 pin */
+ } else {
+ ecs = ecard_find(0, mfm_cids);
+ if (!ecs) {
+ mfm_addr = 0;
+ return -1;
+ }
+
+ mfm_addr = ecard_address(ecs, ECARD_IOC, ECARD_MEDIUM) + 0x800;
+ mfm_IRQPollLoc = mfm_addr + 0x400;
+ mfm_irqenable = mfm_IRQPollLoc;
+ mfm_irq = ecs->irq;
+ irqmask = 0x08;
+
+ ecard_claim(ecs);
+ }
+
+ printk("mfm: found at address %08X, interrupt %d\n", mfm_addr, mfm_irq);
+ request_region (mfm_addr, 10, "mfm");
+
+ /* Stuff for the assembler routines to get to */
+ hdc63463_baseaddress = ioaddr(mfm_addr);
+ hdc63463_irqpolladdress = ioaddr(mfm_IRQPollLoc);
+ hdc63463_irqpollmask = irqmask;
+
+ blk_dev[MAJOR_NR].request_fn = DEVICE_REQUEST;
+ read_ahead[MAJOR_NR] = 8; /* 8 sector (4kB?) read ahread */
+
+#ifndef MODULE
+ mfm_gendisk.next = gendisk_head;
+ gendisk_head = &mfm_gendisk;
+#endif
+
+ Busy = 0;
+ lastspecifieddrive = -1;
+
+ return 0;
+}
+
+/*
+ * This routine is called to flush all partitions and partition tables
+ * for a changed MFM disk, and then re-read the new partition table.
+ * If we are revalidating due to an ioctl, we have USAGE == 1.
+ */
+static int mfm_reread_partitions(kdev_t dev)
+{
+ unsigned int start, i, maxp, target = DEVICE_NR(MINOR(dev));
+ unsigned long flags;
+
+ save_flags_cli(flags);
+ if (mfm_info[target].busy || mfm_info[target].access_count > 1) {
+ restore_flags (flags);
+ return -EBUSY;
+ }
+ mfm_info[target].busy = 1;
+ restore_flags (flags);
+
+ maxp = mfm_gendisk.max_p;
+ start = target << mfm_gendisk.minor_shift;
+
+ for (i = maxp - 1; i >= 0; i--) {
+ int minor = start + i;
+ kdev_t devi = MKDEV(MAJOR_NR, minor);
+ struct super_block *sb = get_super(devi);
+
+ sync_dev (devi);
+ if (sb)
+ invalidate_inodes (sb);
+ invalidate_buffers (devi);
+
+ mfm_gendisk.part[minor].start_sect = 0;
+ mfm_gendisk.part[minor].nr_sects = 0;
+ }
+
+ mfm_gendisk.part[start].nr_sects = mfm_info[target].heads *
+ mfm_info[target].cylinders * mfm_info[target].sectors / 2;
+
+ resetup_one_dev(&mfm_gendisk, target);
+
+ mfm_info[target].busy = 0;
+ wake_up (&mfm_wait_open);
+ return 0;
+}
+
+#ifdef MODULE
+int init_module(void)
+{
+ int ret;
+ ret = mfm_init();
+ if (!ret)
+ mfm_geninit(&mfm_gendisk);
+ return ret;
+}
+
+void cleanup_module(void)
+{
+ if (ecs && mfm_irqenable)
+ outw (0, mfm_irqenable); /* Required to enable IRQs from MFM podule */
+ free_irq(mfm_irq, NULL);
+ unregister_blkdev(MAJOR_NR, "mfm");
+ if (ecs)
+ ecard_release(ecs);
+ if (mfm_addr)
+ release_region(mfm_addr, 10);
+}
+#endif
FUNET's LINUX-ADM group, linux-adm@nic.funet.fi
TCL-scripts by Sam Shen, slshen@lbl.gov