[PATCH] proc: revert /proc/uptime to ->read_proc hook

[PATCH] proc: revert /proc/uptime to ->read_proc hook

[Alexey Dobriyan]

Petr: please confirm.

Al, Christoph: could we please find a way to distinguish reads that
come from pread(2) from reads that come from read(2).

Seqfiles are easily made capable of pread(2) just it's impossible to
figure out in ->read() hook. Otherwise getting rid of ->read_proc
is impossible.


commit 6c87df37dcb9c6c33923707fa5191e0a65874d60
Author: Alexey Dobriyan <adobriyan@gmail.com>
Date:   Mon Oct 27 22:38:27 2008 +0300

    proc: revert /proc/uptime to ->read_proc hook
    
    Turned out some VMware userspace does pread(2) on /proc/uptime, but
    seqfiles currently don't allow pread() resulting in -ESPIPE.
    
    Seqfiles in theory can do pread(), but this can be a long story,
    so revert to ->read_proc until then.
    
    http://bugzilla.kernel.org/show_bug.cgi?id=11856
    
    Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>

diff --git a/fs/proc/uptime.c b/fs/proc/uptime.c
index 0c10a0b..df26aa8 100644
--- a/fs/proc/uptime.c
+++ b/fs/proc/uptime.c
@@ -1,43 +1,45 @@
-#include <linux/fs.h>
 #include <linux/init.h>
 #include <linux/proc_fs.h>
 #include <linux/sched.h>
-#include <linux/seq_file.h>
 #include <linux/time.h>
 #include <asm/cputime.h>
 
-static int uptime_proc_show(struct seq_file *m, void *v)
+static int proc_calc_metrics(char *page, char **start, off_t off,
+				 int count, int *eof, int len)
+{
+	if (len <= off + count)
+		*eof = 1;
+	*start = page + off;
+	len -= off;
+	if (len > count)
+		len = count;
+	if (len < 0)
+		len = 0;
+	return len;
+}
+
+static int uptime_read_proc(char *page, char **start, off_t off, int count,
+			    int *eof, void *data)
 {
 	struct timespec uptime;
 	struct timespec idle;
+	int len;
 	cputime_t idletime = cputime_add(init_task.utime, init_task.stime);
 
 	do_posix_clock_monotonic_gettime(&uptime);
 	monotonic_to_bootbased(&uptime);
 	cputime_to_timespec(idletime, &idle);
-	seq_printf(m, "%lu.%02lu %lu.%02lu\n",
+	len = sprintf(page, "%lu.%02lu %lu.%02lu\n",
 			(unsigned long) uptime.tv_sec,
 			(uptime.tv_nsec / (NSEC_PER_SEC / 100)),
 			(unsigned long) idle.tv_sec,
 			(idle.tv_nsec / (NSEC_PER_SEC / 100)));
-	return 0;
+	return proc_calc_metrics(page, start, off, count, eof, len);
 }
 
-static int uptime_proc_open(struct inode *inode, struct file *file)
-{
-	return single_open(file, uptime_proc_show, NULL);
-}
-
-static const struct file_operations uptime_proc_fops = {
-	.open		= uptime_proc_open,
-	.read		= seq_read,
-	.llseek		= seq_lseek,
-	.release	= single_release,
-};
-
 static int __init proc_uptime_init(void)
 {
-	proc_create("uptime", 0, NULL, &uptime_proc_fops);
+	create_proc_read_entry("uptime", 0, NULL, uptime_read_proc, NULL);
 	return 0;
 }
 module_init(proc_uptime_init);

Re: [PATCH] proc: revert /proc/uptime to ->read_proc hook

[Miklos Szeredi]

On Mon, 27 Oct 2008, Alexey Dobriyan wrote:
> Petr: please confirm.
> 
> Al, Christoph: could we please find a way to distinguish reads that
> come from pread(2) from reads that come from read(2).

You can compare the offset with file->f_pos, and if not equal, do an
lseek, read, lseek-back.  Or more efficiently, save-state, lseek,
read, restore-state.

Miklos