Em Tue, Aug 03, 2010 at 12:48:35PM +0100, Dave Martin escreveu:
quoted text
> Fix buggy-looking code which unnecessarily adjusts the file offset
> fields read from /proc/*/maps.
> 
> This may have gone unnoticed since the offset is usually 0 (and the
> logic in util/symbol.c may work incorrectly for other offset values).
> 
> I make assumptions about the intended design here.  The cover note
> accompanying this patch contains a more detailed explanation.
> 
> Signed-off-by: Dave Martin <dave.martin@linaro.org>

Doing some investigation here...

4af8b35d (Anton Blanchard 2010-04-03 164)      pbf += 3;
4af8b35d (Anton Blanchard 2010-04-03 165)      n = hex2u64(pbf, &vm_pgoff);
4af8b35d (Anton Blanchard 2010-04-03 166)      /* pgoff is in bytes, not pages */
4af8b35d (Anton Blanchard 2010-04-03 167)      if (n >= 0)
4af8b35d (Anton Blanchard 2010-04-03 168)              ev.mmap.pgoff = vm_pgoff << getpagesize();
4af8b35d (Anton Blanchard 2010-04-03 169)      else
4af8b35d (Anton Blanchard 2010-04-03 170)              ev.mmap.pgoff = 0;

commit 4af8b35db6634dd1e0d616de689582b6c93550af
Author: Anton Blanchard <anton@samba.org>
Date:   Sat Apr 3 22:53:31 2010 +1100

    perf symbols: Fill in pgoff in mmap synthesized events
    
    When we synthesize mmap events we need to fill in the pgoff field.
    
    I wasn't able to test this completely since I couldn't find an
    executable region with a non 0 offset. We will see it when we start
    doing data profiling.

------------------------

Yeah, not reassuring comment, looking at fs/proc/task_mmu.c we see:

static void show_map_vma(struct seq_file *m, struct vm_area_struct *vma)
{
        struct mm_struct *mm = vma->vm_mm;
        struct file *file = vma->vm_file;
        int flags = vma->vm_flags;
        unsigned long ino = 0;
        unsigned long long pgoff = 0;
        dev_t dev = 0;
        int len;

        if (file) {
                struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
                dev = inode->i_sb->s_dev;
                ino = inode->i_ino;
                pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
        }

        seq_printf(m, "%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu %n",
                        vma->vm_start,
                        vma->vm_end,
                        flags & VM_READ ? 'r' : '-',
                        flags & VM_WRITE ? 'w' : '-',
                        flags & VM_EXEC ? 'x' : '-',
                        flags & VM_MAYSHARE ? 's' : 'p',
                        pgoff,
                        MAJOR(dev), MINOR(dev), ino, &len);

----------------------------------------------------------------------------

So yes, we're double shifting that, your patch is correct, applying it.

quoted text
> ---
>  tools/perf/util/event.c |    8 +-------
>  1 files changed, 1 insertions(+), 7 deletions(-)
> 
> diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
> index 6b0db55..db8a1d4 100644
> --- a/tools/perf/util/event.c
> +++ b/tools/perf/util/event.c
> @@ -151,7 +151,6 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
>  			continue;
>  		pbf += n + 3;
>  		if (*pbf == 'x') { /* vm_exec */
> -			u64 vm_pgoff;
>  			char *execname = strchr(bf, '/');
>  
>  			/* Catch VDSO */
> @@ -162,12 +161,7 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
>  				continue;
>  
>  			pbf += 3;
> -			n = hex2u64(pbf, &vm_pgoff);
> -			/* pgoff is in bytes, not pages */
> -			if (n >= 0)
> -				ev.mmap.pgoff = vm_pgoff << getpagesize();
> -			else
> -				ev.mmap.pgoff = 0;
> +			n = hex2u64(pbf, &ev.mmap.pgoff);
>  
>  			size = strlen(execname);
>  			execname[size - 1] = '__PLACEHOLDER__1_'; /* Remove \n */
> -- 
> 1.7.0.4
--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/