Commit 2141ec62 authored by Simon Horman's avatar Simon Horman Committed by Greg Kroah-Hartman

Staging: dt3155: remove compat code

Remove compatibility code as this is not an older version of the kernel.
Signed-off-by: default avatarSimon Horman <horms@verge.net.au>
Cc: Scott Smedley <ss@aao.gov.au>
Signed-off-by: default avatarGreg Kroah-Hartman <gregkh@suse.de>
parent cf137d5c
......@@ -54,11 +54,10 @@
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/mm.h> /* PAGE_ALIGN() */
#include <linux/io.h>
#include <asm/page.h>
#include "sysdep.h"
/*#define ALL_DEBUG*/
#define ALL_MSG "allocator: "
......
......@@ -59,13 +59,10 @@ extern void printques(int);
#ifdef MODULE
#include <linux/module.h>
#include <linux/version.h>
#include <linux/interrupt.h>
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,10)
MODULE_LICENSE("GPL");
#endif
#endif
......@@ -105,11 +102,7 @@ int dt3155_errno = 0;
#endif
/* wait queue for interrupts */
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,3,1)
wait_queue_head_t dt3155_read_wait_queue[ MAXBOARDS ];
#else
struct wait_queue *dt3155_read_wait_queue[ MAXBOARDS ];
#endif
#define DT_3155_SUCCESS 0
#define DT_3155_FAILURE -EIO
......@@ -186,11 +179,7 @@ static inline void dt3155_isr( int irq, void *dev_id, struct pt_regs *regs )
{
int minor = -1;
int index;
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
unsigned long flags;
#else
int flags;
#endif
u_long buffer_addr;
/* find out who issued the interrupt */
......@@ -264,21 +253,12 @@ static inline void dt3155_isr( int irq, void *dev_id, struct pt_regs *regs )
buffer_addr = dt3155_fbuffer[ minor ]->
frame_info[ dt3155_fbuffer[ minor ]->active_buf ].addr
+ (DT3155_MAX_ROWS / 2) * stride;
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
local_save_flags(flags);
local_irq_disable();
#else
save_flags( flags );
cli();
#endif
wake_up_interruptible( &dt3155_read_wait_queue[ minor ] );
/* Set up the DMA address for the next field */
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
local_irq_restore(flags);
#else
restore_flags( flags );
#endif
WriteMReg((dt3155_lbase[ minor ] + ODD_DMA_START), buffer_addr);
}
......@@ -350,13 +330,8 @@ static inline void dt3155_isr( int irq, void *dev_id, struct pt_regs *regs )
DT3155_STATE_FLD )
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
local_save_flags(flags);
local_irq_disable();
#else
save_flags( flags );
cli();
#endif
#ifdef DEBUG_QUES_B
printques( minor );
......@@ -412,11 +387,7 @@ static inline void dt3155_isr( int irq, void *dev_id, struct pt_regs *regs )
wake_up_interruptible( &dt3155_read_wait_queue[ minor ] );
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
local_irq_restore(flags);
#else
restore_flags( flags );
#endif
}
......@@ -675,11 +646,6 @@ static int dt3155_mmap (struct file * file, struct vm_area_struct * vma)
/* which device are we mmapping? */
int minor = MINOR(file->f_dentry->d_inode->i_rdev);
unsigned long offset;
/* not actually sure when vm_area_struct changed,
but it was in 2.3 sometime */
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,3,20)
offset = vma->vm_pgoff << PAGE_SHIFT;
if (offset >= __pa(high_memory) || (file->f_flags & O_SYNC))
......@@ -688,34 +654,17 @@ static int dt3155_mmap (struct file * file, struct vm_area_struct * vma)
/* Don't try to swap out physical pages.. */
vma->vm_flags |= VM_RESERVED;
#else
if (vma->vm_offset & ~PAGE_MASK)
return -ENXIO;
offset = vma->vm_offset;
#endif
/* they are mapping the registers or the buffer */
if ((offset == dt3155_status[minor].reg_addr &&
vma->vm_end - vma->vm_start == PCI_PAGE_SIZE) ||
(offset == dt3155_status[minor].mem_addr &&
vma->vm_end - vma->vm_start == dt3155_status[minor].mem_size))
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
if (remap_pfn_range(vma,
vma->vm_start,
offset >> PAGE_SHIFT,
vma->vm_end - vma->vm_start,
vma->vm_page_prot))
#else
if (remap_page_range(vma->vm_start,
offset,
vma->vm_end - vma->vm_start,
vma->vm_page_prot))
#endif
{
vma->vm_page_prot)) {
printk("DT3155: remap_page_range() failed.\n");
return -EAGAIN;
}
......@@ -761,9 +710,6 @@ static int dt3155_open( struct inode* inode, struct file* filep)
printk("DT3155: Device opened.\n");
dt3155_dev_open[ minor ] = 1 ;
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0)
MOD_INC_USE_COUNT;
#endif
dt3155_flush( minor );
......@@ -771,11 +717,7 @@ static int dt3155_open( struct inode* inode, struct file* filep)
int_csr_r.reg = 0;
WriteMReg( (dt3155_lbase[ minor ] + INT_CSR), int_csr_r.reg );
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,3,1)
init_waitqueue_head(&(dt3155_read_wait_queue[minor]));
#else
dt3155_read_wait_queue[minor] = NULL;
#endif
return 0;
}
......@@ -798,9 +740,6 @@ static int dt3155_close( struct inode *inode, struct file *filep)
}
else
{
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0)
MOD_DEC_USE_COUNT;
#endif
dt3155_dev_open[ minor ] = 0;
if (dt3155_status[ minor ].state != DT3155_STATE_IDLE)
......@@ -846,8 +785,6 @@ static int dt3155_read (
/* if (dt3155_status[minor].state == DT3155_STATE_IDLE)*/
/* return -EBUSY;*/
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,3,1)
/* non-blocking reads should return if no data */
if (filep->f_flags & O_NDELAY)
{
......@@ -877,34 +814,6 @@ static int dt3155_read (
}
}
#else
while ((frame_index = dt3155_get_ready_buffer(minor)) < 0 )
{
int index;
if (filep->f_flags & O_NDELAY)
return 0;
/* sleep till data arrives , or we get interrupted */
interruptible_sleep_on(&dt3155_read_wait_queue[minor]);
for (index = 0; index < _NSIG_WORDS; index++)
{
/*
* Changing the next line of code below to this:
* if (current->pending.signal.sig[index] &
* ~current->blocked.sig[index])
* would also work on a 2.4 kernel, however, the above
* method is preferred & more portable.
*/
if (current->signal.sig[index] & ~current->blocked.sig[index])
{
printk ("DT3155: read: interrupted\n");
return -EINTR;
}
}
}
#endif
frame_info_p = &dt3155_status[minor].fbuffer.frame_info[frame_index];
/* make this an offset */
......@@ -1046,12 +955,6 @@ int init_module(void)
int rcode = 0;
char *devname[ MAXBOARDS ];
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,1)
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0)
SET_MODULE_OWNER(&dt3155_fops);
#endif
#endif
devname[ 0 ] = "dt3155a";
#if MAXBOARDS == 2
devname[ 1 ] = "dt3155b";
......@@ -1091,17 +994,6 @@ int init_module(void)
/* Now let's find the hardware. find_PCI() will set ndevices to the
* number of cards found in this machine. */
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0)
if ( !(pcibios_present()) )
{
printk("DT3155: Error: No PCI bios on this machine \n");
if( unregister_chrdev( dt3155_major, "dt3155" ) != 0 )
printk("DT3155: cleanup_module failed\n");
return DT_3155_FAILURE;
}
else
#endif
{
if ( (rcode = find_PCI()) != DT_3155_SUCCESS )
{
......@@ -1199,10 +1091,5 @@ void cleanup_module(void)
free_irq( dt3155_status[ index ].irq, (void*)&dt3155_status[index] );
}
}
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0)
if (MOD_IN_USE)
printk("DT3155: device busy, remove delayed\n");
#endif
}
......@@ -33,13 +33,8 @@ extern u_char *dt3155_bbase;
#ifdef __KERNEL__
#include <linux/wait.h>
#include <linux/version.h> /* need access to LINUX_VERSION_CODE */
/* wait queue for reads */
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 3, 1)
extern wait_queue_head_t dt3155_read_wait_queue[MAXBOARDS];
#else
extern struct wait_queue *dt3155_read_wait_queue[MAXBOARDS];
#endif
#endif
/* number of devices */
......
......@@ -435,20 +435,11 @@ static inline void internal_release_locked_buffer( int m )
*****************************************************/
inline void dt3155_release_locked_buffer( int m )
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
unsigned long int flags;
local_save_flags(flags);
local_irq_disable();
internal_release_locked_buffer(m);
local_irq_restore(flags);
#else
int flags;
save_flags( flags );
cli();
internal_release_locked_buffer( m );
restore_flags( flags );
#endif
}
......@@ -460,15 +451,9 @@ inline void dt3155_release_locked_buffer( int m )
inline int dt3155_flush( int m )
{
int index;
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
unsigned long int flags;
local_save_flags(flags);
local_irq_disable();
#else
int flags;
save_flags( flags );
cli();
#endif
internal_release_locked_buffer( m );
dt3155_fbuffer[ m ]->empty_len = 0;
......@@ -486,11 +471,7 @@ inline int dt3155_flush( int m )
dt3155_fbuffer[ m ]->ready_head = 0;
dt3155_fbuffer[ m ]->ready_len = 0;
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
local_irq_restore(flags);
#else
restore_flags( flags );
#endif
return 0;
}
......@@ -507,15 +488,9 @@ inline int dt3155_flush( int m )
inline int dt3155_get_ready_buffer( int m )
{
int frame_index;
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
unsigned long int flags;
local_save_flags(flags);
local_irq_disable();
#else
int flags;
save_flags( flags );
cli();
#endif
#ifdef DEBUG_QUES_A
printques( m );
......@@ -535,11 +510,7 @@ inline int dt3155_get_ready_buffer( int m )
printques( m );
#endif
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
local_irq_restore(flags);
#else
restore_flags( flags );
#endif
return frame_index;
}
/* This header only makes send when included in a 2.0 compile */
#ifndef _PCI_COMPAT_H_
#define _PCI_COMPAT_H_
#ifdef __KERNEL__
#include <linux/bios32.h> /* pcibios_* */
#include <linux/pci.h> /* pcibios_* */
#include <linux/malloc.h> /* kmalloc */
/* fake the new pci interface based on the old one: encapsulate bus/devfn */
struct pci_fake_dev {
u8 bus;
u8 devfn;
int index;
};
#define pci_dev pci_fake_dev /* the other pci_dev is unused by 2.0 drivers */
extern inline struct pci_dev *pci_find_device(unsigned int vendorid,
unsigned int devid,
struct pci_dev *from)
{
struct pci_dev *pptr = kmalloc(sizeof(*pptr), GFP_KERNEL);
int index = 0;
int ret;
if (!pptr) return NULL;
if (from) index = pptr->index + 1;
ret = pcibios_find_device(vendorid, devid, index,
&pptr->bus, &pptr->devfn);
if (ret) { kfree(pptr); return NULL; }
return pptr;
}
extern inline struct pci_dev *pci_find_class(unsigned int class,
struct pci_dev *from)
{
return NULL; /* FIXME */
}
extern inline void pci_release_device(struct pci_dev *dev)
{
kfree(dev);
}
/* struct pci_dev *pci_find_slot (unsigned int bus, unsigned int devfn); */
#define pci_present pcibios_present
extern inline int
pci_read_config_byte(struct pci_dev *dev, u8 where, u8 *val)
{
return pcibios_read_config_byte(dev->bus, dev->devfn, where, val);
}
extern inline int
pci_read_config_word(struct pci_dev *dev, u8 where, u16 *val)
{
return pcibios_read_config_word(dev->bus, dev->devfn, where, val);
}
extern inline int
pci_read_config_dword(struct pci_dev *dev, u8 where, u32 *val)
{
return pcibios_read_config_dword(dev->bus, dev->devfn, where, val);
}
extern inline int
pci_write_config_byte(struct pci_dev *dev, u8 where, u8 val)
{
return pcibios_write_config_byte(dev->bus, dev->devfn, where, val);
}
extern inline int
pci_write_config_word(struct pci_dev *dev, u8 where, u16 val)
{
return pcibios_write_config_word(dev->bus, dev->devfn, where, val);
}
extern inline int
pci_write_config_dword(struct pci_dev *dev, u8 where, u32 val)
{
return pcibios_write_config_dword(dev->bus, dev->devfn, where, val);
}
extern inline void pci_set_master(struct pci_dev *dev)
{
u16 cmd;
pcibios_read_config_word(dev->bus, dev->devfn, PCI_COMMAND, &cmd);
cmd |= PCI_COMMAND_MASTER;
pcibios_write_config_word(dev->bus, dev->devfn, PCI_COMMAND, cmd);
}
#endif /* __KERNEL__ */
#endif /* _PCI_COMPAT_H_ */
/*
* sysdep.h -- centralizing compatibility issues between 2.0, 2.2, 2.4
*
* Copyright (C) 2001 Alessandro Rubini and Jonathan Corbet
* Copyright (C) 2001 O'Reilly & Associates
*
* The source code in this file can be freely used, adapted,
* and redistributed in source or binary form, so long as an
* acknowledgment appears in derived source files. The citation
* should list that the code comes from the book "Linux Device
* Drivers" by Alessandro Rubini and Jonathan Corbet, published
* by O'Reilly & Associates. No warranty is attached;
* we cannot take responsibility for errors or fitness for use.
*
* $Id: sysdep.h,v 1.2 2005/08/09 06:08:51 ssmedley Exp $
*/
#ifndef _SYSDEP_H_
#define _SYSDEP_H_
#ifndef LINUX_VERSION_CODE
# include <linux/version.h>
#endif
#ifndef KERNEL_VERSION /* pre-2.1.90 didn't have it */
# define KERNEL_VERSION(vers,rel,seq) ( ((vers)<<16) | ((rel)<<8) | (seq) )
#endif
/* only allow 2.0.x 2.2.y and 2.4.z */
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,0,0) /* not < 2.0 */
# error "This kernel is too old: not supported by this file"
#endif
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,7,0) /* not > 2.7, by now */
# error "This kernel is too recent: not supported by this file"
#endif
#if (LINUX_VERSION_CODE & 0xff00) == 1 /* not 2.1 */
# error "Please don't use linux-2.1, use 2.2, 2.4 or 2.6 instead"
#endif
#if (LINUX_VERSION_CODE & 0xff00) == 3 /* not 2.3 */
# error "Please don't use linux-2.3, use 2.4 or 2.6 instead"
#endif
/* remember about the current version */
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,1,0)
# define LINUX_20
#elif LINUX_VERSION_CODE < KERNEL_VERSION(2,3,0)
# define LINUX_22
#else
# define LINUX_24
#endif
/* we can't support versioning in pre-2.4 because we #define some functions */
#if !defined(LINUX_24) && defined(CONFIG_MODVERSIONS)
# error "This sysdep.h can't support CONFIG_MODVERSIONS"
# error "and old kernels at the same time."
# error "Either use 2.4 or avoid using versioning"
#endif
#ifndef LINUX_20 /* include vmalloc.h if this is 2.2/2.4 */
# ifdef VM_READ /* a typical flag defined by mm.h */
# include <linux/vmalloc.h>
# endif
#endif
#include <linux/sched.h>
/* Modularization issues */
#ifdef LINUX_20
# define __USE_OLD_SYMTAB__
# define EXPORT_NO_SYMBOLS register_symtab(NULL);
# define REGISTER_SYMTAB(tab) register_symtab(tab)
#else
# define REGISTER_SYMTAB(tab) /* nothing */
#endif
#ifdef __USE_OLD_SYMTAB__
# define __MODULE_STRING(s) /* nothing */
# define MODULE_PARM(v,t) /* nothing */
# define MODULE_PARM_DESC(v,t) /* nothing */
# define MODULE_AUTHOR(n) /* nothing */
# define MODULE_DESCRIPTION(d) /* nothing */
# define MODULE_SUPPORTED_DEVICE(n) /* nothing */
#endif
/*
* In version 2.2 (up to 2.2.19, at least), the macro for request_module()
* when no kmod is there is wrong. It's a "do {} while 0" but it shouldbe int
*/
#ifdef LINUX_22
# ifndef CONFIG_KMOD
# undef request_module
# define request_module(name) -ENOSYS
# endif
#endif
#ifndef LINUX_20
# include <linux/init.h> /* module_init/module_exit */
#endif
#ifndef module_init
# define module_init(x) int init_module(void) { return x(); }
# define module_exit(x) void cleanup_module(void) { x(); }
#endif
#ifndef SET_MODULE_OWNER
# define SET_MODULE_OWNER(structure) /* nothing */
#endif
/*
* "select" changed in 2.1.23. The implementation is twin, but this
* header is new
*
*/
#ifdef LINUX_20
# define __USE_OLD_SELECT__
#else
# include <linux/poll.h>
#endif
#ifdef LINUX_20
# define INODE_FROM_F(filp) ((filp)->f_inode)
#else
# define INODE_FROM_F(filp) ((filp)->f_dentry->d_inode)
#endif
/* Other changes in the fops are solved using wrappers */
/*
* Wait queues changed with 2.3
*/
#ifndef DECLARE_WAIT_QUEUE_HEAD
# define DECLARE_WAIT_QUEUE_HEAD(head) struct wait_queue *head = NULL
typedef struct wait_queue *wait_queue_head_t;
# define init_waitqueue_head(head) (*(head)) = NULL
/* offer wake_up_sync as an alias for wake_up */
# define wake_up_sync(head) wake_up(head)
# define wake_up_interruptible_sync(head) wake_up_interruptible(head)
/* Pretend we have add_wait_queue_exclusive */
# define add_wait_queue_exclusive(q,entry) add_wait_queue ((q), (entry))
#endif /* no DECLARE_WAIT_QUEUE_HEAD */
/*
* Define wait_event for 2.0 kernels. (This ripped off directly from
* the 2.2.18 sched.h)
*/
#ifdef LINUX_20
#define __wait_event(wq, condition) \
do { \
struct wait_queue __wait; \
\
__wait.task = current; \
add_wait_queue(&wq, &__wait); \
for (;;) { \
current->state = TASK_UNINTERRUPTIBLE; \
mb(); \
if (condition) \
break; \
schedule(); \
} \
current->state = TASK_RUNNING; \
remove_wait_queue(&wq, &__wait); \
} while (0)
#define wait_event(wq, condition) \
do { \
if (condition) \
break; \
__wait_event(wq, condition); \
} while (0)
#define __wait_event_interruptible(wq, condition, ret) \
do { \
struct wait_queue __wait; \
\
__wait.task = current; \
add_wait_queue(&wq, &__wait); \
for (;;) { \
current->state = TASK_INTERRUPTIBLE; \
mb(); \
if (condition) \
break; \
if (!signal_pending(current)) { \
schedule(); \
continue; \
} \
ret = -ERESTARTSYS; \
break; \
} \
current->state = TASK_RUNNING; \
remove_wait_queue(&wq, &__wait); \
} while (0)
#define wait_event_interruptible(wq, condition) \
({ \
int __ret = 0; \
if (!(condition)) \
__wait_event_interruptible(wq, condition, __ret); \
__ret; \
})
#endif
/*
* 2.3 added tasklets
*/
#ifdef LINUX_24
# define HAVE_TASKLETS
#endif
/* FIXME: implement the other versions of wake_up etc */
/*
* access to user space: use the 2.2 functions,
* and implement them as macros for 2.0
*/
#ifdef LINUX_20
# include <asm/segment.h>
# define access_ok(t,a,sz) (verify_area((t),(void *) (a),(sz)) ? 0 : 1)
# define verify_area_20 verify_area
# define copy_to_user(t,f,n) (memcpy_tofs((t), (f), (n)), 0)
# define copy_from_user(t,f,n) (memcpy_fromfs((t), (f), (n)), 0)
# define __copy_to_user(t,f,n) copy_to_user((t), (f), (n))
# define __copy_from_user(t,f,n) copy_from_user((t), (f), (n))
# define PUT_USER(val,add) (put_user((val),(add)), 0)
# define __PUT_USER(val,add) PUT_USER((val),(add))
# define GET_USER(dest,add) ((dest)=get_user((add)), 0)
# define __GET_USER(dest,add) GET_USER((dest),(add))
#else
# include <asm/uaccess.h>
# include <asm/io.h>
# define verify_area_20(t,a,sz) (0) /* == success */
# define PUT_USER put_user
# define __PUT_USER __put_user
# define GET_USER get_user
# define __GET_USER __get_user
#endif
/*
* Allocation issues
*/
#ifdef GFP_USER /* only if mm.h has been included */
# ifdef LINUX_20
# define __GFP_DMA GFP_DMA /* 2.0 didn't have the leading __ */
# endif
# ifndef LINUX_24
# define __GFP_HIGHMEM 0 /* was not there */
# define GFP_HIGHUSER 0 /* idem */
# endif
# ifdef LINUX_20
# define __get_free_pages(a,b) __get_free_pages((a),(b),0)
# endif
# ifndef LINUX_24
# define get_zeroed_page get_free_page
# endif
#endif
/* ioremap */
#if defined(LINUX_20) && defined(_LINUX_MM_H)
# define ioremap_nocache ioremap
# ifndef __i386__
/* This simple approach works for non-PC platforms. */
# define ioremap vremap
# define iounmap vfree
# else /* the PC has <expletive> ISA; 2.2 and 2.4 remap it, 2.0 needs not */
extern inline void *ioremap(unsigned long phys_addr, unsigned long size)
{
if (phys_addr >= 0xA0000 && phys_addr + size <= 0x100000)
return (void *)phys_addr;
return vremap(phys_addr, size);
}
extern inline void iounmap(void *addr)
{
if ((unsigned long)addr >= 0xA0000
&& (unsigned long)addr < 0x100000)
return;
vfree(addr);
}
# endif
#endif
/* Also, define check_mem_region etc */
#ifndef LINUX_24
# define check_mem_region(a,b) 0 /* success */
# define request_mem_region(a,b,c) /* nothing */
# define release_mem_region(a,b) /* nothing */
#endif
/* implement capable() for 2.0 */
#ifdef LINUX_20
# define capable(anything) suser()
#endif
/* The use_count of exec_domain and binfmt changed in 2.1.23 */
#ifdef LINUX_20
# define INCRCOUNT(p) ((p)->module ? __MOD_INC_USE_COUNT((p)->module) : 0)
# define DECRCOUNT(p) ((p)->module ? __MOD_DEC_USE_COUNT((p)->module) : 0)
# define CURRCOUNT(p) ((p)->module && (p)->module->usecount)
#else
# define INCRCOUNT(p) ((p)->use_count++)
# define DECRCOUNT(p) ((p)->use_count--)
# define CURRCOUNT(p) ((p)->use_count)
#endif
/*
* /proc has changed a lot across the versions...
*/
#ifdef LINUX_20
# define USE_PROC_REGISTER
#endif
/*
* 2.2 didn't have create_proc_{read|info}_entry yet.
* And it looks like there are no other "interesting" entry point, as
* the rest is somehow esotique (mknod, symlink, ...)
*/
#ifdef LINUX_22
# ifdef PROC_SUPER_MAGIC /* Only if procfs is being used */
extern inline struct proc_dir_entry *create_proc_read_entry(const char *name,
mode_t mode, struct proc_dir_entry *base,
read_proc_t *read_proc, void * data)
{
struct proc_dir_entry *res=create_proc_entry(name,mode,base);
if (res) {
res->read_proc=read_proc;
res->data=data;
}
return res;
}
# ifndef create_proc_info_entry /* added in 2.2.18 */
typedef int (get_info_t)(char *, char **, off_t, int, int);
extern inline struct proc_dir_entry *create_proc_info_entry(const char *name,
mode_t mode, struct proc_dir_entry *base, get_info_t *get_info)
{
struct proc_dir_entry *res=create_proc_entry(name,mode,base);
if (res) res->get_info=get_info;
return res;
}
# endif /* no create_proc_info_entry */
# endif
#endif
#ifdef LINUX_20
# define test_and_set_bit(nr,addr) test_bit((nr),(addr))
# define test_and_clear_bit(nr,addr) clear_bit((nr),(addr))
# define test_and_change_bit(nr,addr) change_bit((nr),(addr))
#endif
/* 2.0 had no read and write memory barriers, and 2.2 lacks the
set_ functions */
#ifndef LINUX_24
# ifdef LINUX_20
# define wmb() mb() /* this is a big penalty on non-reordering platfs */
# define rmb() mb() /* this is a big penalty on non-reordering platfs */
# endif /* LINUX_20 */
#define set_mb() do { var = value; mb(); } while (0)
#define set_wmb() do { var = value; wmb(); } while (0)
#endif /* ! LINUX_24 */
/* 2.1.30 removed these functions. Let's define them, just in case */
#ifndef LINUX_20
# define queue_task_irq queue_task
# define queue_task_irq_off queue_task
#endif
/* 2.1.10 and 2.1.43 introduced new functions. They are worth using */
#ifdef LINUX_20
# include <asm/byteorder.h>
# ifdef __LITTLE_ENDIAN
# define cpu_to_le16(x) (x)
# define cpu_to_le32(x) (x)
# define cpu_to_be16(x) htons((x))
# define cpu_to_be32(x) htonl((x))
# else
# define cpu_to_be16(x) (x)
# define cpu_to_be32(x) (x)
extern inline __u16 cpu_to_le16(__u16 x) { return (x<<8) | (x>>8);}
extern inline __u32 cpu_to_le32(__u32 x) { return (x>>24) |
((x>>8)&0xff00) | ((x<<8)&0xff0000) | (x<<24);}
# endif
# define le16_to_cpu(x) cpu_to_le16(x)
# define le32_to_cpu(x) cpu_to_le32(x)
# define be16_to_cpu(x) cpu_to_be16(x)
# define be32_to_cpu(x) cpu_to_be32(x)
# define cpu_to_le16p(addr) (cpu_to_le16(*(addr)))
# define cpu_to_le32p(addr) (cpu_to_le32(*(addr)))
# define cpu_to_be16p(addr) (cpu_to_be16(*(addr)))
# define cpu_to_be32p(addr) (cpu_to_be32(*(addr)))
extern inline void cpu_to_le16s(__u16 *a) {*a = cpu_to_le16(*a);}
extern inline void cpu_to_le32s(__u16 *a) {*a = cpu_to_le32(*a);}
extern inline void cpu_to_be16s(__u16 *a) {*a = cpu_to_be16(*a);}
extern inline void cpu_to_be32s(__u16 *a) {*a = cpu_to_be32(*a);}
# define le16_to_cpup(x) cpu_to_le16p(x)
# define le32_to_cpup(x) cpu_to_le32p(x)
# define be16_to_cpup(x) cpu_to_be16p(x)
# define be32_to_cpup(x) cpu_to_be32p(x)
# define le16_to_cpus(x) cpu_to_le16s(x)
# define le32_to_cpus(x) cpu_to_le32s(x)
# define be16_to_cpus(x) cpu_to_be16s(x)
# define be32_to_cpus(x) cpu_to_be32s(x)
#endif
#ifdef LINUX_20
# define __USE_OLD_REBUILD_HEADER__
#endif
/*
* 2.0 didn't include sema_init, so we make our own - but only if it
* looks like semaphore.h got included.
*/
#ifdef LINUX_20
# ifdef MUTEX_LOCKED /* Only if semaphore.h included */
extern inline void sema_init (struct semaphore *sem, int val)
{
sem->count = val;
sem->waking = sem->lock = 0;
sem->wait = NULL;
}
# endif
#endif /* LINUX_20 */
/*
* In 2.0, there is no real need for spinlocks, and they weren't really
* implemented anyway.
*
* XXX the _irqsave variant should be defined eventually to do the
* right thing.
*/
#ifdef LINUX_20
typedef int spinlock_t;
# define spin_lock(lock)
# define spin_unlock(lock)
# define spin_lock_init(lock)
# define spin_lock_irqsave(lock,flags) do { \
save_flags(flags); cli(); } while (0);
# define spin_unlock_irqrestore(lock,flags) restore_flags(flags);
#endif
/*
* 2.1 stuffed the "flush" method into the middle of the file_operations
* structure. The FOP_NO_FLUSH symbol is for drivers that do not implement
* flush (most of them), it can be inserted in initializers for all 2.x
* kernel versions.
*/
#ifdef LINUX_20
# define FOP_NO_FLUSH /* nothing */
# define TAG_LLSEEK lseek
# define TAG_POLL select
#else
# define FOP_NO_FLUSH NULL,
# define TAG_LLSEEK llseek
# define TAG_POLL poll
#endif
/*
* fasync changed in 2.2.
*/
#ifdef LINUX_20
/* typedef struct inode *fasync_file; */
# define fasync_file struct inode *
#else
typedef int fasync_file;
#endif
/* kill_fasync had less arguments, and a different indirection in the first */
#ifndef LINUX_24
# define kill_fasync(ptrptr,sig,band) kill_fasync(*(ptrptr),(sig))
#endif
/* other things that are virtualized: define the new functions for the old k */
#ifdef LINUX_20
# define in_interrupt() (intr_count!=0)
# define mdelay(x) udelay((x)*1000)
# define signal_pending(current) ((current)->signal & ~(current)->blocked)
#endif
#ifdef LINUX_PCI_H /* only if PCI stuff is being used */
# ifdef LINUX_20
# include "pci-compat.h" /* a whole set of replacement functions */
# else
# define pci_release_device(d) /* placeholder, used in 2.0 to free stuff */
# endif
#endif
/*
* Some task state stuff
*/
#ifndef set_current_state
# define set_current_state(s) current->state = (s);
#endif
#ifdef LINUX_20
extern inline void schedule_timeout(int timeout)
{
current->timeout = jiffies + timeout;
current->state = TASK_INTERRUPTIBLE;
schedule();
current->timeout = 0;
}
extern inline long sleep_on_timeout(wait_queue_head_t *q, signed long timeout)
{
signed long early = 0;
current->timeout = jiffies + timeout;
sleep_on (q);
if (current->timeout > 0) {
early = current->timeout - jiffies;
current->timeout = 0;
}
return early;
}
extern inline long interruptible_sleep_on_timeout(wait_queue_head_t *q,
signed long timeout)
{
signed long early = 0;
current->timeout = jiffies + timeout;
interruptible_sleep_on (q);
if (current->timeout > 0) {
early = current->timeout - jiffies;
current->timeout = 0;
}
return early;
}
#endif /* LINUX_20 */
/*
* Schedule_task was a late 2.4 addition.
*/
#ifndef LINUX_24
extern inline int schedule_task(struct tq_struct *task)
{
queue_task(task, &tq_scheduler);
return 1;
}
#endif
/*
* Timing issues
*/
#ifdef LINUX_20
# define get_fast_time do_gettimeofday
#endif
#ifdef _LINUX_DELAY_H /* only if linux/delay.h is included */
# ifndef mdelay /* linux-2.0 */
# ifndef MAX_UDELAY_MS
# define MAX_UDELAY_MS 5
# endif
# define mdelay(n) (\
(__builtin_constant_p(n) && (n)<=MAX_UDELAY_MS) ? udelay((n)*1000) : \
({unsigned long msec=(n); while (msec--) udelay(1000);}))
# endif /* mdelay */
#endif /* _LINUX_DELAY_H */
/*
* No del_timer_sync before 2.4
*/
#ifndef LINUX_24
# define del_timer_sync(timer) del_timer(timer) /* and hope */
#endif
/*
* mod_timer wasn't present in 2.0
*/
#ifdef LINUX_20
static inline int mod_timer(struct timer_list *timer, unsigned long expires)
{
int pending = del_timer(timer);
if (pending) {
timer->expires = expires;
add_timer(timer);
}
return pending;
}
#endif
/*
* Various changes in mmap and friends.
*/
#ifndef NOPAGE_SIGBUS
# define NOPAGE_SIGBUS NULL /* return value of the nopage memory method */
# define NOPAGE_OOM NULL /* No real equivalent in older kernels */
#endif
#ifndef VM_RESERVED /* Added 2.4.0-test10 */
# define VM_RESERVED 0
#endif
#ifdef LINUX_24 /* use "vm_pgoff" to get an offset */
#define VMA_OFFSET(vma) ((vma)->vm_pgoff << PAGE_SHIFT)
#else /* use "vm_offset" */
#define VMA_OFFSET(vma) ((vma)->vm_offset)
#endif
#ifdef MAP_NR
#define virt_to_page(page) (mem_map + MAP_NR(page))
#endif
#ifndef get_page
# define get_page(p) atomic_inc(&(p)->count)
#endif
/*
* No DMA lock in 2.0.
*/
#ifdef LINUX_20
static inline unsigned long claim_dma_lock(void)
{
unsigned long flags;
save_flags(flags);
cli();
return flags;
}
static inline void release_dma_lock(unsigned long flags)
{
restore_flags(flags);
}
#endif
/*
* I/O memory was not managed by ealier kernels, define them as success
*/
#if 0 /* FIXME: what is the right way to do request_mem_region? */
#ifndef LINUX_24
# define check_mem_region(start, len) 0
# define request_mem_region(start, len, name) 0
# define release_mem_region(start, len) 0
/*
* Also, request_ and release_ region used to return void. Return 0 instead
*/
# define request_region(s, l, n) ({request_region((s),(l),(n));0;})
# define release_region(s, l) ({release_region((s),(l));0;})
#endif /* not LINUX_24 */
#endif
/*
* Block layer stuff.
*/
#ifndef LINUX_24
/* BLK_DEFAULT_QUEUE for use with these macros only!!!! */
#define BLK_DEFAULT_QUEUE(major) blk_dev[(major)].request_fn
#define blk_init_queue(where,request_fn) where = request_fn;
#define blk_cleanup_queue(where) where = NULL;
/* No QUEUE_EMPTY in older kernels */
#ifndef QUEUE_EMPTY /* Driver can redefine it too */
# define QUEUE_EMPTY (CURRENT != NULL)
#endif
#ifdef RO_IOCTLS
static inline int blk_ioctl(kdev_t dev, unsigned int cmd, unsigned long arg)
{
int err;
switch (cmd) {
case BLKRAGET: /* return the readahead value */
if (!arg) return -EINVAL;
err = ! access_ok(VERIFY_WRITE, arg, sizeof(long));
if (err) return -EFAULT;
PUT_USER(read_ahead[MAJOR(dev)],(long *) arg);
return 0;
case BLKRASET: /* set the readahead value */
if (!capable(CAP_SYS_ADMIN)) return -EACCES;
if (arg > 0xff) return -EINVAL; /* limit it */
read_ahead[MAJOR(dev)] = arg;
return 0;
case BLKFLSBUF: /* flush */
if (! capable(CAP_SYS_ADMIN)) return -EACCES; /* only root */
fsync_dev(dev);
invalidate_buffers(dev);
return 0;
RO_IOCTLS(dev, arg);
}
return -ENOTTY;
}
#endif /* RO_IOCTLS */
#ifdef LINUX_EXTENDED_PARTITION /* defined in genhd.h */
static inline void register_disk(struct gendisk *gdev, kdev_t dev,
unsigned minors, struct file_operations *ops, long size)
{
if (! gdev)
return;
resetup_one_dev(gdev, MINOR(dev) >> gdev->minor_shift);
}
#endif /* LINUX_EXTENDED_PARTITION */
#else /* it is Linux 2.4 */
#define HAVE_BLKPG_H
#endif /* LINUX_24 */
#ifdef LINUX_20 /* physical and virtual addresses had the same value */
# define __pa(a) (a)
# define __va(a) (a)
#endif
/*
* Network driver compatibility
*/
/*
* 2.0 dev_kfree_skb had an extra arg. The following is a little dangerous
* in that it assumes that FREE_WRITE is always wanted. Very few 2.0 drivers
* use FREE_READ, but the number is *not* zero...
*
* Also: implement the non-checking versions of a couple skb functions -
* but they still check in 2.0.
*/
#ifdef LINUX_20
# define dev_kfree_skb(skb) dev_kfree_skb((skb), FREE_WRITE);
# define __skb_push(skb, len) skb_push((skb), (len))
# define __skb_put(skb, len) skb_put((skb), (len))
#endif
/*
* Softnet changes in 2.4
*/
#ifndef LINUX_24
# ifdef _LINUX_NETDEVICE_H /* only if netdevice.h was included */
# define netif_start_queue(dev) clear_bit(0, (void *) &(dev)->tbusy);
# define netif_stop_queue(dev) set_bit(0, (void *) &(dev)->tbusy);
static inline void netif_wake_queue(struct device *dev)
{
clear_bit(0, (void *) &(dev)->tbusy);
mark_bh(NET_BH);
}
/* struct device became struct net_device */
# define net_device device
# endif /* netdevice.h */
#endif /* ! LINUX_24 */
/*
* Memory barrier stuff, define what's missing from older kernel versions
*/
#ifdef switch_to /* this is always a macro, defined in <asm/sysstem.h> */
# ifndef set_mb
# define set_mb(var, value) do {(var) = (value); mb();} while 0
# endif
# ifndef set_rmb
# define set_rmb(var, value) do {(var) = (value); rmb();} while 0
# endif
# ifndef set_wmb
# define set_wmb(var, value) do {(var) = (value); wmb();} while 0
# endif
/* The hw barriers are defined as sw barriers. A correct thing if this
specific kernel/platform is supported but has no specific instruction */
# ifndef mb
# define mb barrier
# endif
# ifndef rmb
# define rmb barrier
# endif
# ifndef wmb
# define wmb barrier
# endif
#endif /* switch to (i.e. <asm/system.h>) */
#endif /* _SYSDEP_H_ */
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment