zoukankan      html  css  js  c++  java
  • mm/swap

    /*
     *  linux/mm/swap.c
     *
     *  Copyright (C) 1991, 1992  Linus Torvalds
     */

    /*
     * This file should contain most things doing the swapping from/to disk.
     * Started 18.12.91
     */

    #include <linux/mm.h>
    #include <linux/sched.h>
    #include <linux/head.h>
    #include <linux/kernel.h>
    #include <linux/kernel_stat.h>
    #include <linux/errno.h>
    #include <linux/string.h>
    #include <linux/stat.h>

    #include <asm/system.h> /* for cli()/sti() */
    #include <asm/bitops.h>

    #define MAX_SWAPFILES 8

    #define SWP_USED    1
    #define SWP_WRITEOK    3

    #define SWP_TYPE(entry) (((entry) & 0xfe) >> 1)
    #define SWP_OFFSET(entry) ((entry) >> PAGE_SHIFT)
    #define SWP_ENTRY(type,offset) (((type) << 1) | ((offset) << PAGE_SHIFT))

    static int nr_swapfiles = 0;
    static struct wait_queue * lock_queue = NULL;

    //交换信息结构
    static struct swap_info_struct {
        unsigned long flags;
        struct inode * swap_file;
        unsigned int swap_device;
        unsigned char * swap_map;
        unsigned char * swap_lockmap;
        int pages;
        int lowest_bit;
        int highest_bit;
        unsigned long max;
    } swap_info[MAX_SWAPFILES];

    extern unsigned long free_page_list;
    extern int shm_swap (int);

    /*
     * The following are used to make sure we don't thrash too much...
     * NOTE!! NR_LAST_FREE_PAGES must be a power of 2...
     */
    #define NR_LAST_FREE_PAGES 32
    static unsigned long last_free_pages[NR_LAST_FREE_PAGES] = {0,};

    //读写交换页
    void rw_swap_page(int rw, unsigned long entry, char * buf)
    {
        unsigned long type, offset;
        struct swap_info_struct * p;

        type = SWP_TYPE(entry);
        if (type >= nr_swapfiles) {
            printk("Internal error: bad swap-device ");
            return;
        }
        p = &swap_info[type];
        offset = SWP_OFFSET(entry);
        if (offset >= p->max) {
            printk("rw_swap_page: weirdness ");
            return;
        }
        if (!(p->flags & SWP_USED)) {
            printk("Trying to swap to unused swap-device ");
            return;
        }
        while (set_bit(offset,p->swap_lockmap))
            sleep_on(&lock_queue);
        if (rw == READ)
            kstat.pswpin++;
        else
            kstat.pswpout++;
        if (p->swap_device) {
            ll_rw_page(rw,p->swap_device,offset,buf);
        } else if (p->swap_file) {
            unsigned int zones[8];
            unsigned int block;
            int i, j;

            block = offset << (12 - p->swap_file->i_sb->s_blocksize_bits);

            for (i=0, j=0; j< PAGE_SIZE ; i++, j +=p->swap_file->i_sb->s_blocksize)
                if (!(zones[i] = bmap(p->swap_file,block++))) {
                    printk("rw_swap_page: bad swap file ");
                    return;
                }
            ll_rw_swap_file(rw,p->swap_file->i_dev, zones, i,buf);
        } else
            printk("re_swap_page: no swap file or device ");
        if (offset && !clear_bit(offset,p->swap_lockmap))
            printk("rw_swap_page: lock already cleared ");
        wake_up(&lock_queue);
    }

    //获取交换页
    unsigned int get_swap_page(void)
    {
        struct swap_info_struct * p;
        unsigned int offset, type;

        p = swap_info;
        for (type = 0 ; type < nr_swapfiles ; type++,p++) {
            if ((p->flags & SWP_WRITEOK) != SWP_WRITEOK)
                continue;
            for (offset = p->lowest_bit; offset <= p->highest_bit ; offset++) {
                if (p->swap_map[offset])
                    continue;
                p->swap_map[offset] = 1;
                nr_swap_pages--;
                if (offset == p->highest_bit)
                    p->highest_bit--;
                p->lowest_bit = offset;
                return SWP_ENTRY(type,offset);
            }
        }
        return 0;
    }

    //复制交换页
    unsigned long swap_duplicate(unsigned long entry)
    {
        struct swap_info_struct * p;
        unsigned long offset, type;

        if (!entry)
            return 0;
        offset = SWP_OFFSET(entry);
        type = SWP_TYPE(entry);
        if (type == SHM_SWP_TYPE)
            return entry;
        if (type >= nr_swapfiles) {
            printk("Trying to duplicate nonexistent swap-page ");
            return 0;
        }
        p = type + swap_info;
        if (offset >= p->max) {
            printk("swap_free: weirdness ");
            return 0;
        }
        if (!p->swap_map[offset]) {
            printk("swap_duplicate: trying to duplicate unused page ");
            return 0;
        }
        p->swap_map[offset]++;
        return entry;
    }

    //释放
    void swap_free(unsigned long entry)
    {
        struct swap_info_struct * p;
        unsigned long offset, type;

        if (!entry)
            return;
        type = SWP_TYPE(entry);
        if (type == SHM_SWP_TYPE)
            return;
        if (type >= nr_swapfiles) {
            printk("Trying to free nonexistent swap-page ");
            return;
        }
        p = & swap_info[type];
        offset = SWP_OFFSET(entry);
        if (offset >= p->max) {
            printk("swap_free: weirdness ");
            return;
        }
        if (!(p->flags & SWP_USED)) {
            printk("Trying to free swap from unused swap-device ");
            return;
        }
        while (set_bit(offset,p->swap_lockmap))
            sleep_on(&lock_queue);
        if (offset < p->lowest_bit)
            p->lowest_bit = offset;
        if (offset > p->highest_bit)
            p->highest_bit = offset;
        if (!p->swap_map[offset])
            printk("swap_free: swap-space map bad (entry %08lx) ",entry);
        else
            if (!--p->swap_map[offset])
                nr_swap_pages++;
        if (!clear_bit(offset,p->swap_lockmap))
            printk("swap_free: lock already cleared ");
        wake_up(&lock_queue);
    }

    //交换入
    void swap_in(unsigned long *table_ptr)
    {
        unsigned long entry;
        unsigned long page;

        entry = *table_ptr;
        if (PAGE_PRESENT & entry) {
            printk("trying to swap in present page ");
            return;
        }
        if (!entry) {
            printk("No swap page in swap_in ");
            return;
        }
        if (SWP_TYPE(entry) == SHM_SWP_TYPE) {
            shm_no_page ((unsigned long *) table_ptr);
            return;
        }
        if (!(page = get_free_page(GFP_KERNEL))) {
            oom(current);
            page = BAD_PAGE;
        } else    
            read_swap_page(entry, (char *) page);
        if (*table_ptr != entry) {
            free_page(page);
            return;
        }
        *table_ptr = page | (PAGE_DIRTY | PAGE_PRIVATE);
        swap_free(entry);
    }

    //试图交换出
    static inline int try_to_swap_out(unsigned long * table_ptr)
    {
        int i;
        unsigned long page;
        unsigned long entry;

        page = *table_ptr;
        if (!(PAGE_PRESENT & page))
            return 0;
        if (page >= high_memory)
            return 0;
        if (mem_map[MAP_NR(page)] & MAP_PAGE_RESERVED)
            return 0;
        if (PAGE_ACCESSED & page) {
            *table_ptr &= ~PAGE_ACCESSED;
            return 0;
        }
        for (i = 0; i < NR_LAST_FREE_PAGES; i++)
            if (last_free_pages[i] == (page & PAGE_MASK))
                return 0;
        if (PAGE_DIRTY & page) {
            page &= PAGE_MASK;
            if (mem_map[MAP_NR(page)] != 1)
                return 0;
            if (!(entry = get_swap_page()))
                return 0;
            *table_ptr = entry;
            invalidate();
            write_swap_page(entry, (char *) page);
            free_page(page);
            return 1;
        }
        page &= PAGE_MASK;
        *table_ptr = 0;
        invalidate();
        free_page(page);
        return 1 + mem_map[MAP_NR(page)];
    }

    /*
     * sys_idle() does nothing much: it just searches for likely candidates for
     * swapping out or forgetting about. This speeds up the search when we
     * actually have to swap.
     */
     //空闲
    asmlinkage int sys_idle(void)
    {
        need_resched = 1;
        return 0;
    }

    /*
     * A new implementation of swap_out().  We do not swap complete processes,
     * but only a small number of blocks, before we continue with the next
     * process.  The number of blocks actually swapped is determined on the
     * number of page faults, that this process actually had in the last time,
     * so we won't swap heavily used processes all the time ...
     *
     * Note: the priority argument is a hint on much CPU to waste with the
     *       swap block search, not a hint, of how much blocks to swap with
     *       each process.
     *
     * (C) 1993 Kai Petzke, wpp@marie.physik.tu-berlin.de
     */
    #ifdef NEW_SWAP
    /*
     * These are the miminum and maximum number of pages to swap from one process,
     * before proceeding to the next:
     */
    #define SWAP_MIN    4
    #define SWAP_MAX    32

    /*
     * The actual number of pages to swap is determined as:
     * SWAP_RATIO / (number of recent major page faults)
     */
    #define SWAP_RATIO    128

    //交换出
    static int swap_out(unsigned int priority)
    {
        static int swap_task;
        int table;
        int page;
        long pg_table;
        int loop;
        int counter = NR_TASKS * 2 >> priority;
        struct task_struct *p;

        counter = NR_TASKS * 2 >> priority;
        for(; counter >= 0; counter--, swap_task++) {
        /*
         * Check that swap_task is suitable for swapping.  If not, look for
         * the next suitable process.
         */
        loop = 0;
        while(1) {
            if(swap_task >= NR_TASKS) {
            swap_task = 1;
            if(loop)
                /* all processes are unswappable or already swapped out */
                return 0;
            loop = 1;
            }

            p = task[swap_task];
            if(p && p->swappable && p->rss)
            break;

            swap_task++;
        }

        /*
         * Determine the number of pages to swap from this process.
         */
        if(! p -> swap_cnt) {
            p->dec_flt = (p->dec_flt * 3) / 4 + p->maj_flt - p->old_maj_flt;
            p->old_maj_flt = p->maj_flt;

            if(p->dec_flt >= SWAP_RATIO / SWAP_MIN) {
            p->dec_flt = SWAP_RATIO / SWAP_MIN;
            p->swap_cnt = SWAP_MIN;
            } else if(p->dec_flt <= SWAP_RATIO / SWAP_MAX)
            p->swap_cnt = SWAP_MAX;
            else
            p->swap_cnt = SWAP_RATIO / p->dec_flt;
        }

        /*
         * Go through process' page directory.
         */
        for(table = p->swap_table; table < 1024; table++) {
            pg_table = ((unsigned long *) p->tss.cr3)[table];
            if(pg_table >= high_memory)
                continue;
            if(mem_map[MAP_NR(pg_table)] & MAP_PAGE_RESERVED)
                continue;
            if(!(PAGE_PRESENT & pg_table)) {
                printk("swap_out: bad page-table at pg_dir[%d]: %08lx ",
                    table, pg_table);
                ((unsigned long *) p->tss.cr3)[table] = 0;
                continue;
            }
            pg_table &= 0xfffff000;

            /*
             * Go through this page table.
             */
            for(page = p->swap_page; page < 1024; page++) {
            switch(try_to_swap_out(page + (unsigned long *) pg_table)) {
                case 0:
                break;

                case 1:
                p->rss--;
                /* continue with the following page the next time */
                p->swap_table = table;
                p->swap_page  = page + 1;
                if((--p->swap_cnt) == 0)
                    swap_task++;
                return 1;

                default:
                p->rss--;
                break;
            }
            }

            p->swap_page = 0;
        }

        /*
         * Finish work with this process, if we reached the end of the page
         * directory.  Mark restart from the beginning the next time.
         */
        p->swap_table = 0;
        }
        return 0;
    }

    #else /* old swapping procedure */

    /*
     * Go through the page tables, searching for a user page that
     * we can swap out.
     *
     * We now check that the process is swappable (normally only 'init'
     * is un-swappable), allowing high-priority processes which cannot be
     * swapped out (things like user-level device drivers (Not implemented)).
     */
    static int swap_out(unsigned int priority)
    {
        static int swap_task = 1;
        static int swap_table = 0;
        static int swap_page = 0;
        int counter = NR_TASKS*8;
        int pg_table;
        struct task_struct * p;

        counter >>= priority;
    check_task:
        if (counter-- < 0)
            return 0;
        if (swap_task >= NR_TASKS) {
            swap_task = 1;
            goto check_task;
        }
        p = task[swap_task];
        if (!p || !p->swappable) {
            swap_task++;
            goto check_task;
        }
    check_dir:
        if (swap_table >= PTRS_PER_PAGE) {
            swap_table = 0;
            swap_task++;
            goto check_task;
        }
        pg_table = ((unsigned long *) p->tss.cr3)[swap_table];
        if (pg_table >= high_memory || (mem_map[MAP_NR(pg_table)] & MAP_PAGE_RESERVED)) {
            swap_table++;
            goto check_dir;
        }
        if (!(PAGE_PRESENT & pg_table)) {
            printk("bad page-table at pg_dir[%d]: %08x ",
                swap_table,pg_table);
            ((unsigned long *) p->tss.cr3)[swap_table] = 0;
            swap_table++;
            goto check_dir;
        }
        pg_table &= PAGE_MASK;
    check_table:
        if (swap_page >= PTRS_PER_PAGE) {
            swap_page = 0;
            swap_table++;
            goto check_dir;
        }
        switch (try_to_swap_out(swap_page + (unsigned long *) pg_table)) {
            case 0: break;
            case 1: p->rss--; return 1;
            default: p->rss--;
        }
        swap_page++;
        goto check_table;
    }

    #endif

    //尝试释放页面
    static int try_to_free_page(void)
    {
        int i=6;

        while (i--) {
            if (shrink_buffers(i))
                return 1;
            if (shm_swap(i))
                return 1;
            if (swap_out(i))
                return 1;
        }
        return 0;
    }

    /*
     * Note that this must be atomic, or bad things will happen when
     * pages are requested in interrupts (as malloc can do). Thus the
     * cli/sti's.
     */
     //添加内存队列
    static inline void add_mem_queue(unsigned long addr, unsigned long * queue)
    {
        addr &= PAGE_MASK;
        *(unsigned long *) addr = *queue;
        *queue = addr;
    }

    /*
     * Free_page() adds the page to the free lists. This is optimized for
     * fast normal cases (no error jumps taken normally).
     *
     * The way to optimize jumps for gcc-2.2.2 is to:
     *  - select the "normal" case and put it inside the if () { XXX }
     *  - no else-statements if you can avoid them
     *
     * With the above two rules, you get a straight-line execution path
     * for the normal case, giving better asm-code.
     */
     //释放页面
    void free_page(unsigned long addr)
    {
        if (addr < high_memory) {
            unsigned short * map = mem_map + MAP_NR(addr);

            if (*map) {
                if (!(*map & MAP_PAGE_RESERVED)) {
                    unsigned long flag;

                    save_flags(flag);
                    cli();
                    if (!--*map) {
                        if (nr_secondary_pages < MAX_SECONDARY_PAGES) {
                            add_mem_queue(addr,&secondary_page_list);
                            nr_secondary_pages++;
                            restore_flags(flag);
                            return;
                        }
                        add_mem_queue(addr,&free_page_list);
                        nr_free_pages++;
                    }
                    restore_flags(flag);
                }
                return;
            }
            printk("Trying to free free memory (%08lx): memory probabably corrupted ",addr);
            printk("PC = %08lx ",*(((unsigned long *)&addr)-1));
            return;
        }
    }

    /*
     * This is one ugly macro, but it simplifies checking, and makes
     * this speed-critical place reasonably fast, especially as we have
     * to do things with the interrupt flag etc.
     *
     * Note that this #define is heavily optimized to give fast code
     * for the normal case - the if-statements are ordered so that gcc-2.2.2
     * will make *no* jumps for the normal code. Don't touch unless you
     * know what you are doing.
     */
     //从内存队列中移除
    #define REMOVE_FROM_MEM_QUEUE(queue,nr)
        cli();
        if ((result = queue) != 0) {
            if (!(result & ~PAGE_MASK) && result < high_memory) {
                queue = *(unsigned long *) result;
                if (!mem_map[MAP_NR(result)]) {
                    mem_map[MAP_NR(result)] = 1;
                    nr--;
    last_free_pages[index = (index + 1) & (NR_LAST_FREE_PAGES - 1)] = result;
                    restore_flags(flag);
                    return result;
                }
                printk("Free page %08lx has mem_map = %d ",
                    result,mem_map[MAP_NR(result)]);
            } else
                printk("Result = 0x%08lx - memory map destroyed ", result);
            queue = 0;
            nr = 0;
        } else if (nr) {
            printk(#nr " is %d, but " #queue " is empty ",nr);
            nr = 0;
        }
        restore_flags(flag)

    /*
     * Get physical address of first (actually last :-) free page, and mark it
     * used. If no free pages left, return 0.
     *
     * Note that this is one of the most heavily called functions in the kernel,
     * so it's a bit timing-critical (especially as we have to disable interrupts
     * in it). See the above macro which does most of the work, and which is
     * optimized for a fast normal path of execution.
     */
     //获取空闲页
    unsigned long __get_free_page(int priority)
    {
        extern unsigned long intr_count;
        unsigned long result, flag;
        static unsigned long index = 0;

        /* this routine can be called at interrupt time via
           malloc.  We want to make sure that the critical
           sections of code have interrupts disabled. -RAB
           Is this code reentrant? */

        if (intr_count && priority != GFP_ATOMIC) {
            printk("gfp called nonatomically from interrupt %08lx ",
                ((unsigned long *)&priority)[-1]);
            priority = GFP_ATOMIC;
        }
        save_flags(flag);
    repeat:
        REMOVE_FROM_MEM_QUEUE(free_page_list,nr_free_pages);
        if (priority == GFP_BUFFER)
            return 0;
        if (priority != GFP_ATOMIC)
            if (try_to_free_page())
                goto repeat;
        REMOVE_FROM_MEM_QUEUE(secondary_page_list,nr_secondary_pages);
        return 0;
    }

    /*
     * Trying to stop swapping from a file is fraught with races, so
     * we repeat quite a bit here when we have to pause. swapoff()
     * isn't exactly timing-critical, so who cares?
     */
     //尝试取消使用
    static int try_to_unuse(unsigned int type)
    {
        int nr, pgt, pg;
        unsigned long page, *ppage;
        unsigned long tmp = 0;
        struct task_struct *p;

        nr = 0;
    /*
     * When we have to sleep, we restart the whole algorithm from the same
     * task we stopped in. That at least rids us of all races.
     */
    repeat:
        for (; nr < NR_TASKS ; nr++) {
            p = task[nr];
            if (!p)
                continue;
            for (pgt = 0 ; pgt < PTRS_PER_PAGE ; pgt++) {
                ppage = pgt + ((unsigned long *) p->tss.cr3);
                page = *ppage;
                if (!page)
                    continue;
                if (!(page & PAGE_PRESENT) || (page >= high_memory))
                    continue;
                if (mem_map[MAP_NR(page)] & MAP_PAGE_RESERVED)
                    continue;
                ppage = (unsigned long *) (page & PAGE_MASK);    
                for (pg = 0 ; pg < PTRS_PER_PAGE ; pg++,ppage++) {
                    page = *ppage;
                    if (!page)
                        continue;
                    if (page & PAGE_PRESENT)
                        continue;
                    if (SWP_TYPE(page) != type)
                        continue;
                    if (!tmp) {
                        if (!(tmp = __get_free_page(GFP_KERNEL)))
                            return -ENOMEM;
                        goto repeat;
                    }
                    read_swap_page(page, (char *) tmp);
                    if (*ppage == page) {
                        *ppage = tmp | (PAGE_DIRTY | PAGE_PRIVATE);
                        ++p->rss;
                        swap_free(page);
                        tmp = 0;
                    }
                    goto repeat;
                }
            }
        }
        free_page(tmp);
        return 0;
    }

    //关闭交换
    asmlinkage int sys_swapoff(const char * specialfile)
    {
        struct swap_info_struct * p;
        struct inode * inode;
        unsigned int type;
        int i;

        if (!suser())
            return -EPERM;
        i = namei(specialfile,&inode);
        if (i)
            return i;
        p = swap_info;
        for (type = 0 ; type < nr_swapfiles ; type++,p++) {
            if ((p->flags & SWP_WRITEOK) != SWP_WRITEOK)
                continue;
            if (p->swap_file) {
                if (p->swap_file == inode)
                    break;
            } else {
                if (!S_ISBLK(inode->i_mode))
                    continue;
                if (p->swap_device == inode->i_rdev)
                    break;
            }
        }
        iput(inode);
        if (type >= nr_swapfiles)
            return -EINVAL;
        p->flags = SWP_USED;
        i = try_to_unuse(type);
        if (i) {
            p->flags = SWP_WRITEOK;
            return i;
        }
        nr_swap_pages -= p->pages;
        iput(p->swap_file);
        p->swap_file = NULL;
        p->swap_device = 0;
        vfree(p->swap_map);
        p->swap_map = NULL;
        free_page((long) p->swap_lockmap);
        p->swap_lockmap = NULL;
        p->flags = 0;
        return 0;
    }

    /*
     * Written 01/25/92 by Simmule Turner, heavily changed by Linus.
     *
     * The swapon system call
     */
     //打开交换
    asmlinkage int sys_swapon(const char * specialfile)
    {
        struct swap_info_struct * p;
        struct inode * swap_inode;
        unsigned int type;
        int i,j;
        int error;

        if (!suser())
            return -EPERM;
        p = swap_info;
        for (type = 0 ; type < nr_swapfiles ; type++,p++)
            if (!(p->flags & SWP_USED))
                break;
        if (type >= MAX_SWAPFILES)
            return -EPERM;
        if (type >= nr_swapfiles)
            nr_swapfiles = type+1;
        p->flags = SWP_USED;
        p->swap_file = NULL;
        p->swap_device = 0;
        p->swap_map = NULL;
        p->swap_lockmap = NULL;
        p->lowest_bit = 0;
        p->highest_bit = 0;
        p->max = 1;
        error = namei(specialfile,&swap_inode);
        if (error)
            goto bad_swap;
        error = -EBUSY;
        if (swap_inode->i_count != 1)
            goto bad_swap;
        error = -EINVAL;
        if (S_ISBLK(swap_inode->i_mode)) {
            p->swap_device = swap_inode->i_rdev;
            iput(swap_inode);
            error = -ENODEV;
            if (!p->swap_device)
                goto bad_swap;
            error = -EBUSY;
            for (i = 0 ; i < nr_swapfiles ; i++) {
                if (i == type)
                    continue;
                if (p->swap_device == swap_info[i].swap_device)
                    goto bad_swap;
            }
        } else if (S_ISREG(swap_inode->i_mode))
            p->swap_file = swap_inode;
        else
            goto bad_swap;
        p->swap_lockmap = (unsigned char *) get_free_page(GFP_USER);
        if (!p->swap_lockmap) {
            printk("Unable to start swapping: out of memory :-) ");
            error = -ENOMEM;
            goto bad_swap;
        }
        read_swap_page(SWP_ENTRY(type,0), (char *) p->swap_lockmap);
        if (memcmp("SWAP-SPACE",p->swap_lockmap+4086,10)) {
            printk("Unable to find swap-space signature ");
            error = -EINVAL;
            goto bad_swap;
        }
        memset(p->swap_lockmap+PAGE_SIZE-10,0,10);
        j = 0;
        p->lowest_bit = 0;
        p->highest_bit = 0;
        for (i = 1 ; i < 8*PAGE_SIZE ; i++) {
            if (test_bit(i,p->swap_lockmap)) {
                if (!p->lowest_bit)
                    p->lowest_bit = i;
                p->highest_bit = i;
                p->max = i+1;
                j++;
            }
        }
        if (!j) {
            printk("Empty swap-file ");
            error = -EINVAL;
            goto bad_swap;
        }
        p->swap_map = (unsigned char *) vmalloc(p->max);
        if (!p->swap_map) {
            error = -ENOMEM;
            goto bad_swap;
        }
        for (i = 1 ; i < p->max ; i++) {
            if (test_bit(i,p->swap_lockmap))
                p->swap_map[i] = 0;
            else
                p->swap_map[i] = 0x80;
        }
        p->swap_map[0] = 0x80;
        memset(p->swap_lockmap,0,PAGE_SIZE);
        p->flags = SWP_WRITEOK;
        p->pages = j;
        nr_swap_pages += j;
        printk("Adding Swap: %dk swap-space ",j<<2);
        return 0;
    bad_swap:
        free_page((long) p->swap_lockmap);
        vfree(p->swap_map);
        iput(p->swap_file);
        p->swap_device = 0;
        p->swap_file = NULL;
        p->swap_map = NULL;
        p->swap_lockmap = NULL;
        p->flags = 0;
        return error;
    }
    //交换信息
    void si_swapinfo(struct sysinfo *val)
    {
        unsigned int i, j;

        val->freeswap = val->totalswap = 0;
        for (i = 0; i < nr_swapfiles; i++) {
            if (!(swap_info[i].flags & SWP_USED))
                continue;
            for (j = 0; j < swap_info[i].max; ++j)
                switch (swap_info[i].swap_map[j]) {
                    case 128:
                        continue;
                    case 0:
                        ++val->freeswap;
                    default:
                        ++val->totalswap;
                }
        }
        val->freeswap <<= PAGE_SHIFT;
        val->totalswap <<= PAGE_SHIFT;
        return;
    }

  • 相关阅读:
    Linux 系统中用户切换(su user与 su
    linux 用户打开进程数和文件数调整
    hive sql 语法详解
    iOS
    iOS
    MySQL的事务的处理
    iOS
    iOS AOP编程思想及实践
    iOS 静态库和动态库(库详解)
    iOS 沙盒目录结构及正确使用
  • 原文地址:https://www.cnblogs.com/xiaofengwei/p/3773994.html
Copyright © 2011-2022 走看看