Jeremy Fitzhardinge
2009-Jan-22 22:36 UTC
[Xen-devel] [PATCH] xen: actually release memory when shrinking domain
From: Dan Magenheimer <dan.magenheimer@oracle.com> Fix this:> It appears that in the upstream balloon driver, > > the call to HYPERVISOR_update_va_mapping is missing > > from decrease_reservation. I think as a result, > > the balloon driver is eating memory but not > > releasing it to Xen, thus rendering the balloon > > driver essentially useless. (Can be observed via xentop.)Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com> Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> --- drivers/xen/balloon.c | 5 +++++ 1 file changed, 5 insertions(+) ==================================================================--- linux-2.6.28/drivers/xen/balloon.c.orig +++ linux-2.6.28/drivers/xen/balloon.c @@ -296,6 +296,11 @@ frame_list[i] = pfn_to_mfn(pfn); scrub_page(page); + + ret = HYPERVISOR_update_va_mapping( + (unsigned long)__va(pfn << PAGE_SHIFT), + __pte_ma(0), 0); + BUG_ON(ret); } /* Ensure that ballooned highmem pages don''t have kmaps. */ _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ingo Molnar
2009-Jan-22 22:42 UTC
[Xen-devel] Re: [PATCH] xen: actually release memory when shrinking domain
* Jeremy Fitzhardinge <jeremy@goop.org> wrote:> From: Dan Magenheimer <dan.magenheimer@oracle.com> > > Fix this: > >> It appears that in the upstream balloon driver, >> > the call to HYPERVISOR_update_va_mapping is missing >> > from decrease_reservation. I think as a result, >> > the balloon driver is eating memory but not >> > releasing it to Xen, thus rendering the balloon >> > driver essentially useless. (Can be observed via xentop.) > > Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com> > Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> > --- > drivers/xen/balloon.c | 5 +++++ > 1 file changed, 5 insertions(+)applied to tip/x86/urgent, thanks! Ingo _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Campbell
2009-Jan-23 14:09 UTC
Re: [Xen-devel] [PATCH] xen: actually release memory when shrinking domain
On Thu, 2009-01-22 at 14:36 -0800, Jeremy Fitzhardinge wrote:> From: Dan Magenheimer <dan.magenheimer@oracle.com> > > Fix this: > > > It appears that in the upstream balloon driver, > > > the call to HYPERVISOR_update_va_mapping is missing > > > from decrease_reservation. I think as a result, > > > the balloon driver is eating memory but not > > > releasing it to Xen, thus rendering the balloon > > > driver essentially useless. (Can be observed via xentop.)Fails on 32 bit with HighMem pages: kernel BUG at /local/scratch/ianc/devel/kernels/paravirt/drivers/xen/balloon.c:298! invalid opcode: 0000 [#1] SMP last sysfs file: /sys/devices/system/xen_memory/xen_memory0/info/current_kb Modules linked in: Pid: 9, comm: events/0 Tainted: G W (2.6.29-rc1-x86_32p-xen0-tip #726) EIP: 0061:[<c0266c57>] EFLAGS: 00010282 CPU: 0 EIP is at balloon_process+0x527/0x530 EAX: ffffffea EBX: deadbeef ECX: deadbeef EDX: deadbeef ESI: deadbeef EDI: c1000000 EBP: 00000000 ESP: f64b5f44 DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0069 Process events/0 (pid: 9, ti=f64b4000 task=f642d830 task.ti=f64b4000) Stack: 00000000 ffffa0c0 0000002e 00000192 0000019d 00000000 00000400 f6427a00 00000000 00000000 00000000 00000000 00007ff0 c04afcb0 f6427a00 c04afcac c0266730 c0143cf1 c0106bc2 f6427a0c c0107810 f642020c f6427a04 f6427a0c Call Trace: [<c0266730>] balloon_process+0x0/0x530 [<c0143cf1>] run_workqueue+0x91/0x140 [<c0106bc2>] check_events+0x8/0x16 [<c0107810>] xen_spin_unlock+0x0/0x50 [<c01473e0>] autoremove_wake_function+0x0/0x50 [<c0144718>] worker_thread+0x98/0xf0 [<c01473e0>] autoremove_wake_function+0x0/0x50 [<c0144680>] worker_thread+0x0/0xf0 [<c01470c2>] kthread+0x42/0x70 [<c0147080>] kthread+0x0/0x70 [<c010a5f3>] kernel_thread_helper+0x7/0x10 Code: 00 00 00 00 e8 9b a1 f0 ff 31 f6 e8 e4 2f eb ff b8 e0 7b 54 c0 e8 1a 90 16 00 8b 6c 24 18 89 44 24 14 89 74 24 10 e9 eb fe ff ff <0f> 0b eb fe 90 8d 74 26 00 83 ec 04 b8 20 fc 4a c0 e8 13 dc ff EIP: [<c0266c57>] balloon_process+0x527/0x530 SS:ESP 0069:f64b5f44 ---[ end trace 4eaa2a86a8e2da28 ]--- diff -r b64a63f076cd drivers/xen/balloon.c --- a/drivers/xen/balloon.c Fri Jan 23 13:29:48 2009 +0000 +++ b/drivers/xen/balloon.c Fri Jan 23 14:09:08 2009 +0000 @@ -292,10 +292,13 @@ scrub_page(page); - ret = HYPERVISOR_update_va_mapping( - (unsigned long)__va(pfn << PAGE_SHIFT), - __pte_ma(0), 0); - BUG_ON(ret); + if (!PageHighMem(page)) { + ret = HYPERVISOR_update_va_mapping( + (unsigned long)__va(pfn << PAGE_SHIFT), + __pte_ma(0), 0); + BUG_ON(ret); + } + } /* Ensure that ballooned highmem pages don''t have kmaps. */ _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jeremy Fitzhardinge
2009-Jan-23 16:16 UTC
Re: [Xen-devel] [PATCH] xen: actually release memory when shrinking domain
Ian Campbell wrote:> On Thu, 2009-01-22 at 14:36 -0800, Jeremy Fitzhardinge wrote: > >> From: Dan Magenheimer <dan.magenheimer@oracle.com> >> >> Fix this: >> >> >>> It appears that in the upstream balloon driver, >>> >>>> the call to HYPERVISOR_update_va_mapping is missing >>>> from decrease_reservation. I think as a result, >>>> the balloon driver is eating memory but not >>>> releasing it to Xen, thus rendering the balloon >>>> driver essentially useless. (Can be observed via xentop.) >>>> > > Fails on 32 bit with HighMem pages: > > kernel BUG at /local/scratch/ianc/devel/kernels/paravirt/drivers/xen/balloon.c:298! >Ah, good point. Could you package this up into a proper patch for Ingo? Thanks, J> invalid opcode: 0000 [#1] SMP > last sysfs file: /sys/devices/system/xen_memory/xen_memory0/info/current_kb > Modules linked in: > > Pid: 9, comm: events/0 Tainted: G W (2.6.29-rc1-x86_32p-xen0-tip #726) > EIP: 0061:[<c0266c57>] EFLAGS: 00010282 CPU: 0 > EIP is at balloon_process+0x527/0x530 > EAX: ffffffea EBX: deadbeef ECX: deadbeef EDX: deadbeef > ESI: deadbeef EDI: c1000000 EBP: 00000000 ESP: f64b5f44 > DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0069 > Process events/0 (pid: 9, ti=f64b4000 task=f642d830 task.ti=f64b4000) > Stack: > 00000000 ffffa0c0 0000002e 00000192 0000019d 00000000 00000400 f6427a00 > 00000000 00000000 00000000 00000000 00007ff0 c04afcb0 f6427a00 c04afcac > c0266730 c0143cf1 c0106bc2 f6427a0c c0107810 f642020c f6427a04 f6427a0c > Call Trace: > [<c0266730>] balloon_process+0x0/0x530 > [<c0143cf1>] run_workqueue+0x91/0x140 > [<c0106bc2>] check_events+0x8/0x16 > [<c0107810>] xen_spin_unlock+0x0/0x50 > [<c01473e0>] autoremove_wake_function+0x0/0x50 > [<c0144718>] worker_thread+0x98/0xf0 > [<c01473e0>] autoremove_wake_function+0x0/0x50 > [<c0144680>] worker_thread+0x0/0xf0 > [<c01470c2>] kthread+0x42/0x70 > [<c0147080>] kthread+0x0/0x70 > [<c010a5f3>] kernel_thread_helper+0x7/0x10 > Code: 00 00 00 00 e8 9b a1 f0 ff 31 f6 e8 e4 2f eb ff b8 e0 7b 54 c0 e8 1a 90 16 00 8b 6c 24 18 89 44 24 14 89 74 24 10 e9 eb fe ff ff <0f> 0b eb fe 90 8d 74 26 00 83 ec 04 b8 20 fc 4a c0 e8 13 dc ff > EIP: [<c0266c57>] balloon_process+0x527/0x530 SS:ESP 0069:f64b5f44 > ---[ end trace 4eaa2a86a8e2da28 ]--- > > diff -r b64a63f076cd drivers/xen/balloon.c > --- a/drivers/xen/balloon.c Fri Jan 23 13:29:48 2009 +0000 > +++ b/drivers/xen/balloon.c Fri Jan 23 14:09:08 2009 +0000 > @@ -292,10 +292,13 @@ > > scrub_page(page); > > - ret = HYPERVISOR_update_va_mapping( > - (unsigned long)__va(pfn << PAGE_SHIFT), > - __pte_ma(0), 0); > - BUG_ON(ret); > + if (!PageHighMem(page)) { > + ret = HYPERVISOR_update_va_mapping( > + (unsigned long)__va(pfn << PAGE_SHIFT), > + __pte_ma(0), 0); > + BUG_ON(ret); > + } > + > } > > /* Ensure that ballooned highmem pages don''t have kmaps. */ > > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel