Re: [PATCH v3] proc: faster open/read/close with "permanent" files

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Sun, Feb 23, 2020 at 06:48:38PM -0800, Joe Perches wrote:
> On Sun, 2020-02-23 at 14:30 +0300, Alexey Dobriyan wrote:
> > On Sat, Feb 22, 2020 at 12:39:39PM -0800, Joe Perches wrote:
> > > On Sat, 2020-02-22 at 23:15 +0300, Alexey Dobriyan wrote:
> > > > Now that "struct proc_ops" exist we can start putting there stuff which
> > > > could not fly with VFS "struct file_operations"...
> > > > 
> > > > Most of fs/proc/inode.c file is dedicated to make open/read/.../close reliable
> > > > in the event of disappearing /proc entries which usually happens if module is
> > > > getting removed. Files like /proc/cpuinfo which never disappear simply do not
> > > > need such protection.
> > > > 
> > > > Save 2 atomic ops, 1 allocation, 1 free per open/read/close sequence for such
> > > > "permanent" files.
> > > > 
> > > > Enable "permanent" flag for
> > > > 
> > > > 	/proc/cpuinfo
> > > > 	/proc/kmsg
> > > > 	/proc/modules
> > > > 	/proc/slabinfo
> > > > 	/proc/stat
> > > > 	/proc/sysvipc/*
> > > > 	/proc/swaps
> > > > 
> > > > More will come once I figure out foolproof way to prevent out module
> > > > authors from marking their stuff "permanent" for performance reasons
> > > > when it is not.
> > > > 
> > > > This should help with scalability: benchmark is "read /proc/cpuinfo R times
> > > > by N threads scattered over the system".
> > > 
> > > Is this an actual expected use-case?
> > 
> > Yes.
> > 
> > > Is there some additional unnecessary memory consumption
> > > in the unscaled systems?
> > 
> > No, it's the opposite. Less memory usage for everyone and noticeable
> > performance improvement for contented case.
> > 
> > > >  static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> > > >  {
> > > >  	struct proc_dir_entry *pde = PDE(file_inode(file));
> > > >  	ssize_t rv = -EIO;
> > > > -	if (use_pde(pde)) {
> > > > -		typeof_member(struct proc_ops, proc_read) read;
> > > >  
> > > > -		read = pde->proc_ops->proc_read;
> > > > -		if (read)
> > > > -			rv = read(file, buf, count, ppos);
> > > > +	if (pde_is_permanent(pde)) {
> > > > +		return pde_read(pde, file, buf, count, ppos);
> > > > +	} else if (use_pde(pde)) {
> > > > +		rv = pde_read(pde, file, buf, count, ppos);
> > > >  		unuse_pde(pde);
> > > 
> > > Perhaps all the function call duplication could be minimized
> > > by using code without direct returns like:
> > > 
> > > 	rv = pde_read(pde, file, buf, count, pos);
> > > 	if (!pde_is_permanent(pde))
> > > 		unuse_pde(pde);
> > > 
> > > 	return rv;
> > 
> > Function call non-duplication is false goal.
> 
> Depends, copy/paste errors are common and object code
> size generally increases.
> 
> > Surprisingly it makes code bigger:
> 
> Not so far as I can tell.  Are you sure?
> 
> > 	$ ./scripts/bloat-o-meter ../vmlinux-000 ../obj/vmlinux
> > 	add/remove: 0/0 grow/shrink: 1/0 up/down: 10/0 (10)
> > 	Function                                     old     new   delta
> > 	proc_reg_read                                108     118     +10
> > 
> > and worse too: "rv" is carried on stack through "unuse_pde" call.
> 
> With gcc 9.2.1 x86-64 defconfig:
> 
> Changing just proc_reg_read to:
> 
> static ssize_t proc_reg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> {
> 	struct proc_dir_entry *pde = PDE(file_inode(file));
> 	ssize_t rv;
> 
> 	rv = pde_read(pde, file, buf, count, ppos);
> 	if (use_pde(pde))
> 		unuse_pde(pde);

What?

Please make non-racy patch before doing anything.

> 
> 	return rv;
> }



[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]

  Powered by Linux