LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Jeremy Fitzhardinge <jeremy@goop.org>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Andi Kleen <ak@muc.de>,
	linux-kernel@vger.kernel.org, virtualization@lists.osdl.org,
	xen-devel@lists.xensource.com, Chris Wright <chrisw@sous-sol.org>,
	Zachary Amsden <zach@vmware.com>
Subject: Re: [patch 10/21] Xen-paravirt: add hooks to intercept mm creation and destruction
Date: Thu, 15 Feb 2007 22:55:33 -0800	[thread overview]
Message-ID: <45D55565.4020206@goop.org> (raw)
In-Reply-To: <20070215223409.eea8f1e3.akpm@linux-foundation.org>

Andrew Morton wrote:
> On Thu, 15 Feb 2007 18:24:59 -0800 Jeremy Fitzhardinge <jeremy@goop.org> wrote:
>
>   
>> Add hooks to allow a paravirt implementation to track the lifetime of
>> an mm.
>>
>> --- a/arch/i386/kernel/paravirt.c
>> +++ b/arch/i386/kernel/paravirt.c
>> @@ -706,6 +706,10 @@ struct paravirt_ops paravirt_ops = {
>>  	.irq_enable_sysexit = native_irq_enable_sysexit,
>>  	.iret = native_iret,
>>  
>> +	.dup_mmap = (void *)native_nop,
>> +	.exit_mmap = (void *)native_nop,
>> +	.activate_mm = (void *)native_nop,
>> +
>>  	.startup_ipi_hook = (void *)native_nop,
>>  };
>>     
>
> eww.  I suppose there's a good reason for the casting.
>   

Yeah, it's a bit ugly.  The alternative is to have a separate
correctly-typed nop function for each operation.  But that's even more
typing.

> It seems strange to call out to arch_foo() from within an arch header file.
> I mean, we implicity know we're i386.
>
> Maybe it's just poorly named.
>   

The other two are arch_* and are called from common code. 
arch_activate_mm() is either empty or a call to
paravirt_ops.activate_mm.  I could name it paravirt_activate_mm (as it
was in earlier versions of this patch), but then it would be
inconsistent with the other functions.  I thought the consistency was
more important, because these calls need to be properly matched.

>> +static inline void paravirt_activate_mm(struct mm_struct *prev,
>> +					struct mm_struct *next)
>> +{
>> +}
>> +
>> +static inline void paravirt_dup_mmap(struct mm_struct *oldmm,
>> +				     struct mm_struct *mm)
>> +{
>> +}
>> +
>> +static inline void paravirt_exit_mmap(struct mm_struct *mm)
>> +{
>> +}
>>     
>
> These functions are unreferenced in this patchset.
>   

OK, I'll drop them.

>>  #endif /* CONFIG_PARAVIRT */
>>  #endif	/* __ASM_PARAVIRT_H */
>> ===================================================================
>> --- a/include/linux/sched.h
>> +++ b/include/linux/sched.h
>> @@ -374,6 +374,12 @@ struct mm_struct {
>>  	rwlock_t		ioctx_list_lock;
>>  	struct kioctx		*ioctx_list;
>>  };
>> +
>> +#ifndef __HAVE_ARCH_MM_LIFETIME
>> +#define arch_activate_mm(prev, next)	do {} while(0)
>> +#define arch_dup_mmap(oldmm, mm)	do {} while(0)
>> +#define arch_exit_mmap(mm)		do {} while(0)
>> +#endif
>>     
>
> Can we lose __HAVE_ARCH_MM_LIFETIME?  Just define these (preferably in C,
> not in cpp) in the appropriate include/asm-foo/ files?
>   

I guess, if you want.  For everything except i386 (and x86_64 in the not
too distant future) they'll be noops.  But for consistency, I/we would
have to put the appropriate arch_activate_mm() into each arch's
activate_mm(); I seem to remember some were not as straightforward as i386.

>>  struct sighand_struct {
>>  	atomic_t		count;
>> ===================================================================
>> --- a/kernel/fork.c
>> +++ b/kernel/fork.c
>> @@ -286,6 +286,7 @@ static inline int dup_mmap(struct mm_str
>>  		if (retval)
>>  			goto out;
>>  	}
>> +	arch_dup_mmap(oldmm, mm);
>>  	retval = 0;
>>  out:
>>  	up_write(&mm->mmap_sem);
>> ===================================================================
>> --- a/mm/mmap.c
>> +++ b/mm/mmap.c
>> @@ -1976,6 +1976,8 @@ void exit_mmap(struct mm_struct *mm)
>>  	struct vm_area_struct *vma = mm->mmap;
>>  	unsigned long nr_accounted = 0;
>>  	unsigned long end;
>> +
>> +	arch_exit_mmap(mm);
>>  
>>  	lru_add_drain();
>>  	flush_cache_mm(mm);
>>     
>
> Perhaps some commentary telling the arch maintainer what these hooks he's
> being offered are for?
>   

OK.

    J

  reply	other threads:[~2007-02-16  6:55 UTC|newest]

Thread overview: 97+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-02-16  2:24 [patch 00/21] Xen-paravirt: Xen guest implementation for paravirt_ops interface Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 01/21] Xen-paravirt: Fix typo in sync_constant_test_bit()s name Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 02/21] Xen-paravirt: ignore vgacon if hardware not present Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 03/21] Xen-paravirt: Add pagetable accessors to pack and unpack pagetable entries Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 04/21] Xen-paravirt: =================================================================== Jeremy Fitzhardinge
2007-02-16  9:12   ` Andi Kleen
2007-02-16 17:32     ` Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 05/21] Xen-paravirt: paravirt_ops: hooks to set up initial pagetable Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 06/21] Xen-paravirt: paravirt_ops: allocate a fixmap slot Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 07/21] Xen-paravirt: remove ctor for pgd cache Jeremy Fitzhardinge
2007-02-16  6:04   ` Zachary Amsden
2007-02-16  8:39   ` Pekka Enberg
2007-02-16  9:19     ` Nick Piggin
2007-02-16 17:19       ` Jeremy Fitzhardinge
2007-02-16 20:33   ` Christoph Lameter
2007-02-16 23:33     ` Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 08/21] Xen-paravirt: Allow paravirt backend to choose kernel PMD sharing Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 09/21] Xen-paravirt: Allow paravirt backend to select PGD allocation alignment Jeremy Fitzhardinge
2007-02-16  2:24 ` [patch 10/21] Xen-paravirt: add hooks to intercept mm creation and destruction Jeremy Fitzhardinge
2007-02-16  6:34   ` Andrew Morton
2007-02-16  6:55     ` Jeremy Fitzhardinge [this message]
2007-02-22 20:16   ` Pavel Machek
2007-02-22 20:21     ` Jeremy Fitzhardinge
2007-02-22 20:23     ` Jeremy Fitzhardinge
2007-02-16  2:25 ` [patch 11/21] Xen-paravirt: Add apply_to_page_range() which applies a function to a pte range Jeremy Fitzhardinge
2007-02-16  6:37   ` Andrew Morton
2007-02-16  7:06     ` Jeremy Fitzhardinge
2007-02-16  7:19       ` Andrew Morton
2007-02-16  7:23         ` Jeremy Fitzhardinge
2007-02-16  7:31     ` Nick Piggin
2007-02-16  2:25 ` [patch 12/21] Xen-paravirt: Allocate and free vmalloc areas Jeremy Fitzhardinge
2007-02-16  6:43   ` Andrew Morton
2007-02-16  7:08     ` Jeremy Fitzhardinge
2007-02-16  7:24       ` Andrew Morton
2007-02-16  7:30         ` Jeremy Fitzhardinge
2007-02-16  7:49           ` Andrew Morton
2007-02-16  9:18           ` Andi Kleen
2007-02-16 11:10             ` [Xen-devel] " Keir Fraser
2007-02-16 11:34               ` Andi Kleen
2007-02-16 16:46             ` Jeremy Fitzhardinge
2007-02-16 17:10               ` [Xen-devel] " Keir Fraser
2007-02-16 17:12                 ` Keir Fraser
2007-02-16 17:27                   ` Jeremy Fitzhardinge
2007-02-16 19:06                     ` Keir Fraser
2007-02-16 19:19                       ` Keir Fraser
2007-02-16 19:26                         ` Jeremy Fitzhardinge
2007-02-16 23:29                           ` Keir Fraser
2007-02-16 23:41                             ` Jeremy Fitzhardinge
2007-02-16 17:26                 ` Hollis Blanchard
2007-02-16  7:48         ` Nick Piggin
2007-02-16  2:25 ` [patch 13/21] Xen-paravirt: Add nosegneg capability to the vsyscall page notes Jeremy Fitzhardinge
2007-02-16  6:06   ` Zachary Amsden
2007-02-16  2:25 ` [patch 14/21] Xen-paravirt: Add XEN config options and disable unsupported config options Jeremy Fitzhardinge
2007-02-16  6:14   ` Dan Hecht
2007-02-16  7:04     ` Jeremy Fitzhardinge
2007-02-16  7:52       ` Dan Hecht
2007-02-16  8:05         ` Jeremy Fitzhardinge
2007-02-16  8:37           ` Dan Hecht
2007-02-16 10:49             ` Keir Fraser
2007-02-16  7:06     ` Andrew Morton
2007-02-16  7:25       ` Jeremy Fitzhardinge
2007-02-16 10:09         ` Keir Fraser
2007-02-16 10:19           ` [Xen-devel] " Zachary Amsden
2007-02-16 10:44             ` Keir Fraser
2007-02-16 10:55               ` Zachary Amsden
2007-02-17  4:04             ` Rusty Russell
2007-02-16  6:15   ` Zachary Amsden
2007-02-16  7:33     ` Eric W. Biederman
2007-02-16  8:14       ` Ian Campbell
2007-02-16  8:24         ` Eric W. Biederman
2007-02-16  8:31           ` Zachary Amsden
2007-02-16 13:53       ` [Xen-devel] " Gerd Hoffmann
2007-02-18 11:32     ` Avi Kivity
2007-02-16  2:25 ` [patch 16/21] Xen-paravirt: Core Xen implementation Jeremy Fitzhardinge
2007-02-16  2:25 ` [patch 17/21] Xen-paravirt: Add the Xen virtual console driver Jeremy Fitzhardinge
2007-02-16  6:48   ` Andrew Morton
2007-02-16  2:25 ` [patch 18/21] Xen-paravirt: Add Xen grant table support Jeremy Fitzhardinge
2007-02-16  6:52   ` Andrew Morton
2007-02-16  6:58     ` Jeremy Fitzhardinge
2007-02-16  2:25 ` [patch 19/21] Xen-paravirt: Add the Xenbus sysfs and virtual device hotplug driver Jeremy Fitzhardinge
2007-02-16  2:25 ` [patch 20/21] Xen-paravirt: Add Xen virtual block device driver Jeremy Fitzhardinge
2007-02-16  2:25 ` [patch 21/21] Xen-paravirt: Add the Xen virtual network " Jeremy Fitzhardinge
2007-02-16  6:59 ` [patch 00/21] Xen-paravirt: Xen guest implementation for paravirt_ops interface Andrew Morton
2007-02-16  7:20   ` Jeremy Fitzhardinge
2007-02-16 20:49 ` Christoph Lameter
2007-02-16 21:04   ` Zachary Amsden
2007-02-16 21:13     ` Christoph Lameter
2007-02-16 21:48       ` Zachary Amsden
2007-02-16 21:59         ` Christoph Lameter
2007-02-16 22:10           ` Zachary Amsden
2007-02-17 13:51           ` Andi Kleen
2007-02-21 18:37             ` Christoph Lameter
2007-02-21 18:55               ` Zachary Amsden
2007-02-21 20:02               ` Jeremy Fitzhardinge
2007-02-17  5:05         ` Rusty Russell
2007-02-16 23:49   ` Jeremy Fitzhardinge
2007-02-17  4:58   ` Rusty Russell

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=45D55565.4020206@goop.org \
    --to=jeremy@goop.org \
    --cc=ak@muc.de \
    --cc=akpm@linux-foundation.org \
    --cc=chrisw@sous-sol.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=virtualization@lists.osdl.org \
    --cc=xen-devel@lists.xensource.com \
    --cc=zach@vmware.com \
    --subject='Re: [patch 10/21] Xen-paravirt: add hooks to intercept mm creation and destruction' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).