* Re: xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
[not found] ` <46118631-61d4-adb6-6ffc-4e7c62ea3da9@arm.com>
@ 2019-03-12 17:14 ` Matthew Wilcox
2019-03-12 17:18 ` David Hildenbrand
2019-03-12 17:23 ` David Hildenbrand
0 siblings, 2 replies; 16+ messages in thread
From: Matthew Wilcox @ 2019-03-12 17:14 UTC (permalink / raw)
To: Julien Grall
Cc: osstest service owner, xen-devel, Juergen Gross,
Konrad Rzeszutek Wilk, Boris Ostrovsky, Stefano Stabellini,
linux-kernel, Kees Cook, david, k.khlebnikov, Julien Freche,
Nadav Amit, VMware, Inc.,
linux-mm
On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
> On 3/12/19 3:59 PM, Julien Grall wrote:
> > It looks like all the arm test for linus [1] and next [2] tree
> > are now failing. x86 seems to be mostly ok.
> >
> > The bisector fingered the following commit:
> >
> > commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
> > Author: Matthew Wilcox <willy@infradead.org>
> > Date: Tue Mar 5 15:46:06 2019 -0800
> >
> > mm/memory.c: prevent mapping typed pages to userspace
> > Pages which use page_type must never be mapped to userspace as it would
> > destroy their page type. Add an explicit check for this instead of
> > assuming that kernel drivers always get this right.
Oh good, it found a real problem.
> It turns out the problem is because the balloon driver will call
> __SetPageOffline() on allocated page. Therefore the page has a type and
> vm_insert_pages will deny the insertion.
>
> My knowledge is quite limited in this area. So I am not sure how we can
> solve the problem.
>
> I would appreciate if someone could provide input of to fix the mapping.
I don't know the balloon driver, so I don't know why it was doing this,
but what it was doing was Wrong and has been since 2014 with:
commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
Date: Thu Oct 9 15:29:27 2014 -0700
mm/balloon_compaction: redesign ballooned pages management
If ballooned pages are supposed to be mapped into userspace, you can't mark
them as ballooned pages using the mapcount field.
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 17:14 ` xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm) Matthew Wilcox
@ 2019-03-12 17:18 ` David Hildenbrand
2019-03-12 17:24 ` [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS " Andrew Cooper
2019-03-12 17:39 ` xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] " Julien Grall
2019-03-12 17:23 ` David Hildenbrand
1 sibling, 2 replies; 16+ messages in thread
From: David Hildenbrand @ 2019-03-12 17:18 UTC (permalink / raw)
To: Matthew Wilcox, Julien Grall
Cc: osstest service owner, xen-devel, Juergen Gross,
Konrad Rzeszutek Wilk, Boris Ostrovsky, Stefano Stabellini,
linux-kernel, Kees Cook, k.khlebnikov, Julien Freche, Nadav Amit,
VMware, Inc.,
linux-mm
On 12.03.19 18:14, Matthew Wilcox wrote:
> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>> It looks like all the arm test for linus [1] and next [2] tree
>>> are now failing. x86 seems to be mostly ok.
>>>
>>> The bisector fingered the following commit:
>>>
>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>> Author: Matthew Wilcox <willy@infradead.org>
>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>
>>> mm/memory.c: prevent mapping typed pages to userspace
>>> Pages which use page_type must never be mapped to userspace as it would
>>> destroy their page type. Add an explicit check for this instead of
>>> assuming that kernel drivers always get this right.
>
> Oh good, it found a real problem.
>
>> It turns out the problem is because the balloon driver will call
>> __SetPageOffline() on allocated page. Therefore the page has a type and
>> vm_insert_pages will deny the insertion.
>>
>> My knowledge is quite limited in this area. So I am not sure how we can
>> solve the problem.
>>
>> I would appreciate if someone could provide input of to fix the mapping.
>
> I don't know the balloon driver, so I don't know why it was doing this,
> but what it was doing was Wrong and has been since 2014 with:
>
> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
> Date: Thu Oct 9 15:29:27 2014 -0700
>
> mm/balloon_compaction: redesign ballooned pages management
>
> If ballooned pages are supposed to be mapped into userspace, you can't mark
> them as ballooned pages using the mapcount field.
>
Asking myself why anybody would want to map balloon inflated pages into
user space (this just sounds plain wrong but my understanding to what
XEN balloon driver does might be limited), but I assume the easy fix
would be to revert
commit 2f085ff37d08ecbc7849d5abb9424bd7927dda1d
Author: David Hildenbrand <david@redhat.com>
Date: Wed Mar 6 11:42:24 2019 +1100
xen/balloon: mark inflated pages PG_offline
Mark inflated and never onlined pages PG_offline, to tell the world that
the content is stale and should not be dumped.
--
Thanks,
David / dhildenb
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 17:14 ` xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm) Matthew Wilcox
2019-03-12 17:18 ` David Hildenbrand
@ 2019-03-12 17:23 ` David Hildenbrand
2019-03-12 17:25 ` Nadav Amit
1 sibling, 1 reply; 16+ messages in thread
From: David Hildenbrand @ 2019-03-12 17:23 UTC (permalink / raw)
To: Matthew Wilcox, Julien Grall
Cc: osstest service owner, xen-devel, Juergen Gross,
Konrad Rzeszutek Wilk, Boris Ostrovsky, Stefano Stabellini,
linux-kernel, Kees Cook, k.khlebnikov, Julien Freche, Nadav Amit,
VMware, Inc.,
linux-mm
On 12.03.19 18:14, Matthew Wilcox wrote:
> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>> It looks like all the arm test for linus [1] and next [2] tree
>>> are now failing. x86 seems to be mostly ok.
>>>
>>> The bisector fingered the following commit:
>>>
>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>> Author: Matthew Wilcox <willy@infradead.org>
>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>
>>> mm/memory.c: prevent mapping typed pages to userspace
>>> Pages which use page_type must never be mapped to userspace as it would
>>> destroy their page type. Add an explicit check for this instead of
>>> assuming that kernel drivers always get this right.
>
> Oh good, it found a real problem.
>
>> It turns out the problem is because the balloon driver will call
>> __SetPageOffline() on allocated page. Therefore the page has a type and
>> vm_insert_pages will deny the insertion.
>>
>> My knowledge is quite limited in this area. So I am not sure how we can
>> solve the problem.
>>
>> I would appreciate if someone could provide input of to fix the mapping.
>
> I don't know the balloon driver, so I don't know why it was doing this,
> but what it was doing was Wrong and has been since 2014 with:
Just to clarify on that point, XEN balloon does not use balloon
compaction as far as I know (only virtio-balloon and as far as I know
now also vmware balloon). Both of them don't map any such pages to user
space, so it never was and isn't a problem.
>
> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
> Date: Thu Oct 9 15:29:27 2014 -0700
>
> mm/balloon_compaction: redesign ballooned pages management
>
> If ballooned pages are supposed to be mapped into userspace, you can't mark
> them as ballooned pages using the mapcount field.
>
--
Thanks,
David / dhildenb
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 17:18 ` David Hildenbrand
@ 2019-03-12 17:24 ` Andrew Cooper
2019-03-12 18:02 ` Boris Ostrovsky
2019-03-12 17:39 ` xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] " Julien Grall
1 sibling, 1 reply; 16+ messages in thread
From: Andrew Cooper @ 2019-03-12 17:24 UTC (permalink / raw)
To: David Hildenbrand, Matthew Wilcox, Julien Grall
Cc: Juergen Gross, k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel, Boris Ostrovsky
On 12/03/2019 17:18, David Hildenbrand wrote:
> On 12.03.19 18:14, Matthew Wilcox wrote:
>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>> are now failing. x86 seems to be mostly ok.
>>>>
>>>> The bisector fingered the following commit:
>>>>
>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>
>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>> Pages which use page_type must never be mapped to userspace as it would
>>>> destroy their page type. Add an explicit check for this instead of
>>>> assuming that kernel drivers always get this right.
>> Oh good, it found a real problem.
>>
>>> It turns out the problem is because the balloon driver will call
>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>> vm_insert_pages will deny the insertion.
>>>
>>> My knowledge is quite limited in this area. So I am not sure how we can
>>> solve the problem.
>>>
>>> I would appreciate if someone could provide input of to fix the mapping.
>> I don't know the balloon driver, so I don't know why it was doing this,
>> but what it was doing was Wrong and has been since 2014 with:
>>
>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>> Date: Thu Oct 9 15:29:27 2014 -0700
>>
>> mm/balloon_compaction: redesign ballooned pages management
>>
>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>> them as ballooned pages using the mapcount field.
>>
> Asking myself why anybody would want to map balloon inflated pages into
> user space (this just sounds plain wrong but my understanding to what
> XEN balloon driver does might be limited), but I assume the easy fix
> would be to revert
I suspect the bug here is that the balloon driver is (ab)used for a
second purpose - to create a hole in pfn space to map some other bits of
shared memory into.
I think at the end of the day, what is needed is a struct page_info
which looks like normal RAM, but the backing for which can be altered by
hypercall to map other things.
~Andrew
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 17:23 ` David Hildenbrand
@ 2019-03-12 17:25 ` Nadav Amit
0 siblings, 0 replies; 16+ messages in thread
From: Nadav Amit @ 2019-03-12 17:25 UTC (permalink / raw)
To: David Hildenbrand
Cc: Matthew Wilcox, Julien Grall, osstest service owner, xen-devel,
Juergen Gross, Konrad Rzeszutek Wilk, Boris Ostrovsky,
Stefano Stabellini, linux-kernel, Kees Cook, k.khlebnikov,
Julien Freche, Pv-drivers, linux-mm
> On Mar 12, 2019, at 10:23 AM, David Hildenbrand <david@redhat.com> wrote:
>
> On 12.03.19 18:14, Matthew Wilcox wrote:
>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>> are now failing. x86 seems to be mostly ok.
>>>>
>>>> The bisector fingered the following commit:
>>>>
>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>
>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>> Pages which use page_type must never be mapped to userspace as it would
>>>> destroy their page type. Add an explicit check for this instead of
>>>> assuming that kernel drivers always get this right.
>>
>> Oh good, it found a real problem.
>>
>>> It turns out the problem is because the balloon driver will call
>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>> vm_insert_pages will deny the insertion.
>>>
>>> My knowledge is quite limited in this area. So I am not sure how we can
>>> solve the problem.
>>>
>>> I would appreciate if someone could provide input of to fix the mapping.
>>
>> I don't know the balloon driver, so I don't know why it was doing this,
>> but what it was doing was Wrong and has been since 2014 with:
>
> Just to clarify on that point, XEN balloon does not use balloon
> compaction as far as I know (only virtio-balloon and as far as I know
> now also vmware balloon). Both of them don't map any such pages to user
> space, so it never was and isn't a problem.
I still need to submit the next version of the patches, but yes, we are not
about to map them to userspace (dah).
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 17:18 ` David Hildenbrand
2019-03-12 17:24 ` [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS " Andrew Cooper
@ 2019-03-12 17:39 ` Julien Grall
2019-03-12 17:49 ` David Hildenbrand
1 sibling, 1 reply; 16+ messages in thread
From: Julien Grall @ 2019-03-12 17:39 UTC (permalink / raw)
To: David Hildenbrand, Matthew Wilcox
Cc: osstest service owner, xen-devel, Juergen Gross,
Konrad Rzeszutek Wilk, Boris Ostrovsky, Stefano Stabellini,
linux-kernel, Kees Cook, k.khlebnikov, Julien Freche, Nadav Amit,
VMware, Inc.,
linux-mm
Hi David,
On 3/12/19 5:18 PM, David Hildenbrand wrote:
> On 12.03.19 18:14, Matthew Wilcox wrote:
>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>> are now failing. x86 seems to be mostly ok.
>>>>
>>>> The bisector fingered the following commit:
>>>>
>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>
>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>> Pages which use page_type must never be mapped to userspace as it would
>>>> destroy their page type. Add an explicit check for this instead of
>>>> assuming that kernel drivers always get this right.
>>
>> Oh good, it found a real problem.
>>
>>> It turns out the problem is because the balloon driver will call
>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>> vm_insert_pages will deny the insertion.
>>>
>>> My knowledge is quite limited in this area. So I am not sure how we can
>>> solve the problem.
>>>
>>> I would appreciate if someone could provide input of to fix the mapping.
>>
>> I don't know the balloon driver, so I don't know why it was doing this,
>> but what it was doing was Wrong and has been since 2014 with:
>>
>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>> Date: Thu Oct 9 15:29:27 2014 -0700
>>
>> mm/balloon_compaction: redesign ballooned pages management
>>
>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>> them as ballooned pages using the mapcount field.
>>
>
> Asking myself why anybody would want to map balloon inflated pages into
> user space (this just sounds plain wrong but my understanding to what
> XEN balloon driver does might be limited), but I assume the easy fix
> would be to revert
Balloon pages are used to map foreign guest pages. As backend PV drivers
may live in userspace (e.g QEMU, Xenconsoled...) we need to be able to
to insert balloon pages in the VM.
>
>
> commit 2f085ff37d08ecbc7849d5abb9424bd7927dda1d
I guess you meant 77c4adf6a6df6f8f39807eaed48eb73d0eb4261e?
I have reverted the patch and can now access the guest console. Is there
a way to keep this patch and at the same time mapping the page in the
userspace?
> Author: David Hildenbrand <david@redhat.com>
> Date: Wed Mar 6 11:42:24 2019 +1100
>
> xen/balloon: mark inflated pages PG_offline
>
> Mark inflated and never onlined pages PG_offline, to tell the world that
> the content is stale and should not be dumped.
>
>
Cheers,
--
Julien Grall
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 17:39 ` xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] " Julien Grall
@ 2019-03-12 17:49 ` David Hildenbrand
0 siblings, 0 replies; 16+ messages in thread
From: David Hildenbrand @ 2019-03-12 17:49 UTC (permalink / raw)
To: Julien Grall, Matthew Wilcox
Cc: osstest service owner, xen-devel, Juergen Gross,
Konrad Rzeszutek Wilk, Boris Ostrovsky, Stefano Stabellini,
linux-kernel, Kees Cook, k.khlebnikov, Julien Freche, Nadav Amit,
VMware, Inc.,
linux-mm
On 12.03.19 18:39, Julien Grall wrote:
> Hi David,
>
> On 3/12/19 5:18 PM, David Hildenbrand wrote:
>> On 12.03.19 18:14, Matthew Wilcox wrote:
>>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>>> are now failing. x86 seems to be mostly ok.
>>>>>
>>>>> The bisector fingered the following commit:
>>>>>
>>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>>
>>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>>> Pages which use page_type must never be mapped to userspace as it would
>>>>> destroy their page type. Add an explicit check for this instead of
>>>>> assuming that kernel drivers always get this right.
>>>
>>> Oh good, it found a real problem.
>>>
>>>> It turns out the problem is because the balloon driver will call
>>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>>> vm_insert_pages will deny the insertion.
>>>>
>>>> My knowledge is quite limited in this area. So I am not sure how we can
>>>> solve the problem.
>>>>
>>>> I would appreciate if someone could provide input of to fix the mapping.
>>>
>>> I don't know the balloon driver, so I don't know why it was doing this,
>>> but what it was doing was Wrong and has been since 2014 with:
>>>
>>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>>> Date: Thu Oct 9 15:29:27 2014 -0700
>>>
>>> mm/balloon_compaction: redesign ballooned pages management
>>>
>>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>>> them as ballooned pages using the mapcount field.
>>>
>>
>> Asking myself why anybody would want to map balloon inflated pages into
>> user space (this just sounds plain wrong but my understanding to what
>> XEN balloon driver does might be limited), but I assume the easy fix
>> would be to revert
>
> Balloon pages are used to map foreign guest pages. As backend PV drivers
> may live in userspace (e.g QEMU, Xenconsoled...) we need to be able to
> to insert balloon pages in the VM.
Okay, so this is really XEN specific (especially looking at Andrew's
reply). All other balloon drivers told the hypervisor that the inflated
page is dead and it is not going to be use before telling the hypervisor
otherwise. Mapping to user space would violate that contract (and even
be considered harmful in some hypervisor implementations).
>
>>
>>
>> commit 2f085ff37d08ecbc7849d5abb9424bd7927dda1d
>
> I guess you meant 77c4adf6a6df6f8f39807eaed48eb73d0eb4261e?
Yes indeed, no idea where that commit id came from :)
>
> I have reverted the patch and can now access the guest console. Is there
> a way to keep this patch and at the same time mapping the page in the
> userspace?
Not without another page flag. And we all know that is unlikely to happen :)
Thanks!
>
>
>> Author: David Hildenbrand <david@redhat.com>
>> Date: Wed Mar 6 11:42:24 2019 +1100
>>
>> xen/balloon: mark inflated pages PG_offline
>>
>> Mark inflated and never onlined pages PG_offline, to tell the world that
>> the content is stale and should not be dumped.
>>
>>
>
> Cheers,
>
--
Thanks,
David / dhildenb
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 17:24 ` [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS " Andrew Cooper
@ 2019-03-12 18:02 ` Boris Ostrovsky
2019-03-12 18:11 ` Andrew Cooper
2019-03-12 18:23 ` David Hildenbrand
0 siblings, 2 replies; 16+ messages in thread
From: Boris Ostrovsky @ 2019-03-12 18:02 UTC (permalink / raw)
To: Andrew Cooper, David Hildenbrand, Matthew Wilcox, Julien Grall
Cc: Juergen Gross, k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 3/12/19 1:24 PM, Andrew Cooper wrote:
> On 12/03/2019 17:18, David Hildenbrand wrote:
>> On 12.03.19 18:14, Matthew Wilcox wrote:
>>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>>> are now failing. x86 seems to be mostly ok.
>>>>>
>>>>> The bisector fingered the following commit:
>>>>>
>>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>>
>>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>>> Pages which use page_type must never be mapped to userspace as it would
>>>>> destroy their page type. Add an explicit check for this instead of
>>>>> assuming that kernel drivers always get this right.
>>> Oh good, it found a real problem.
>>>
>>>> It turns out the problem is because the balloon driver will call
>>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>>> vm_insert_pages will deny the insertion.
>>>>
>>>> My knowledge is quite limited in this area. So I am not sure how we can
>>>> solve the problem.
>>>>
>>>> I would appreciate if someone could provide input of to fix the mapping.
>>> I don't know the balloon driver, so I don't know why it was doing this,
>>> but what it was doing was Wrong and has been since 2014 with:
>>>
>>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>>> Date: Thu Oct 9 15:29:27 2014 -0700
>>>
>>> mm/balloon_compaction: redesign ballooned pages management
>>>
>>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>>> them as ballooned pages using the mapcount field.
>>>
>> Asking myself why anybody would want to map balloon inflated pages into
>> user space (this just sounds plain wrong but my understanding to what
>> XEN balloon driver does might be limited), but I assume the easy fix
>> would be to revert
> I suspect the bug here is that the balloon driver is (ab)used for a
> second purpose
Yes. And its name is alloc_xenballooned_pages().
-boris
> - to create a hole in pfn space to map some other bits of
> shared memory into.
>
> I think at the end of the day, what is needed is a struct page_info
> which looks like normal RAM, but the backing for which can be altered by
> hypercall to map other things.
>
> ~Andrew
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 18:02 ` Boris Ostrovsky
@ 2019-03-12 18:11 ` Andrew Cooper
2019-03-12 18:23 ` David Hildenbrand
1 sibling, 0 replies; 16+ messages in thread
From: Andrew Cooper @ 2019-03-12 18:11 UTC (permalink / raw)
To: Boris Ostrovsky, David Hildenbrand, Matthew Wilcox, Julien Grall
Cc: Juergen Gross, k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 12/03/2019 18:02, Boris Ostrovsky wrote:
> On 3/12/19 1:24 PM, Andrew Cooper wrote:
>> On 12/03/2019 17:18, David Hildenbrand wrote:
>>> On 12.03.19 18:14, Matthew Wilcox wrote:
>>>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>>>> are now failing. x86 seems to be mostly ok.
>>>>>>
>>>>>> The bisector fingered the following commit:
>>>>>>
>>>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>>>
>>>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>>>> Pages which use page_type must never be mapped to userspace as it would
>>>>>> destroy their page type. Add an explicit check for this instead of
>>>>>> assuming that kernel drivers always get this right.
>>>> Oh good, it found a real problem.
>>>>
>>>>> It turns out the problem is because the balloon driver will call
>>>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>>>> vm_insert_pages will deny the insertion.
>>>>>
>>>>> My knowledge is quite limited in this area. So I am not sure how we can
>>>>> solve the problem.
>>>>>
>>>>> I would appreciate if someone could provide input of to fix the mapping.
>>>> I don't know the balloon driver, so I don't know why it was doing this,
>>>> but what it was doing was Wrong and has been since 2014 with:
>>>>
>>>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>>>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>>>> Date: Thu Oct 9 15:29:27 2014 -0700
>>>>
>>>> mm/balloon_compaction: redesign ballooned pages management
>>>>
>>>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>>>> them as ballooned pages using the mapcount field.
>>>>
>>> Asking myself why anybody would want to map balloon inflated pages into
>>> user space (this just sounds plain wrong but my understanding to what
>>> XEN balloon driver does might be limited), but I assume the easy fix
>>> would be to revert
>> I suspect the bug here is that the balloon driver is (ab)used for a
>> second purpose
> Yes. And its name is alloc_xenballooned_pages().
FWIW, I did express my views that this was a BadIdea(tm) when that logic
was first introduced.
But yes - now is clearly the time to fix this properly.
~Andrew
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 18:02 ` Boris Ostrovsky
2019-03-12 18:11 ` Andrew Cooper
@ 2019-03-12 18:23 ` David Hildenbrand
2019-03-12 19:46 ` David Hildenbrand
1 sibling, 1 reply; 16+ messages in thread
From: David Hildenbrand @ 2019-03-12 18:23 UTC (permalink / raw)
To: Boris Ostrovsky, Andrew Cooper, Matthew Wilcox, Julien Grall
Cc: Juergen Gross, k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 12.03.19 19:02, Boris Ostrovsky wrote:
> On 3/12/19 1:24 PM, Andrew Cooper wrote:
>> On 12/03/2019 17:18, David Hildenbrand wrote:
>>> On 12.03.19 18:14, Matthew Wilcox wrote:
>>>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>>>> are now failing. x86 seems to be mostly ok.
>>>>>>
>>>>>> The bisector fingered the following commit:
>>>>>>
>>>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>>>
>>>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>>>> Pages which use page_type must never be mapped to userspace as it would
>>>>>> destroy their page type. Add an explicit check for this instead of
>>>>>> assuming that kernel drivers always get this right.
>>>> Oh good, it found a real problem.
>>>>
>>>>> It turns out the problem is because the balloon driver will call
>>>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>>>> vm_insert_pages will deny the insertion.
>>>>>
>>>>> My knowledge is quite limited in this area. So I am not sure how we can
>>>>> solve the problem.
>>>>>
>>>>> I would appreciate if someone could provide input of to fix the mapping.
>>>> I don't know the balloon driver, so I don't know why it was doing this,
>>>> but what it was doing was Wrong and has been since 2014 with:
>>>>
>>>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>>>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>>>> Date: Thu Oct 9 15:29:27 2014 -0700
>>>>
>>>> mm/balloon_compaction: redesign ballooned pages management
>>>>
>>>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>>>> them as ballooned pages using the mapcount field.
>>>>
>>> Asking myself why anybody would want to map balloon inflated pages into
>>> user space (this just sounds plain wrong but my understanding to what
>>> XEN balloon driver does might be limited), but I assume the easy fix
>>> would be to revert
>> I suspect the bug here is that the balloon driver is (ab)used for a
>> second purpose
>
> Yes. And its name is alloc_xenballooned_pages().
>
Haven't had a look at the code yet, but would another temporary fix be
to clear/set PG_offline when allocating/freeing a ballooned page?
(assuming here that only such pages will be mapped to user space)
--
Thanks,
David / dhildenb
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 18:23 ` David Hildenbrand
@ 2019-03-12 19:46 ` David Hildenbrand
2019-03-14 8:37 ` Juergen Gross
0 siblings, 1 reply; 16+ messages in thread
From: David Hildenbrand @ 2019-03-12 19:46 UTC (permalink / raw)
To: Boris Ostrovsky, Andrew Cooper, Matthew Wilcox, Julien Grall
Cc: Juergen Gross, k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 12.03.19 19:23, David Hildenbrand wrote:
> On 12.03.19 19:02, Boris Ostrovsky wrote:
>> On 3/12/19 1:24 PM, Andrew Cooper wrote:
>>> On 12/03/2019 17:18, David Hildenbrand wrote:
>>>> On 12.03.19 18:14, Matthew Wilcox wrote:
>>>>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>>>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>>>>> are now failing. x86 seems to be mostly ok.
>>>>>>>
>>>>>>> The bisector fingered the following commit:
>>>>>>>
>>>>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>>>>
>>>>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>>>>> Pages which use page_type must never be mapped to userspace as it would
>>>>>>> destroy their page type. Add an explicit check for this instead of
>>>>>>> assuming that kernel drivers always get this right.
>>>>> Oh good, it found a real problem.
>>>>>
>>>>>> It turns out the problem is because the balloon driver will call
>>>>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>>>>> vm_insert_pages will deny the insertion.
>>>>>>
>>>>>> My knowledge is quite limited in this area. So I am not sure how we can
>>>>>> solve the problem.
>>>>>>
>>>>>> I would appreciate if someone could provide input of to fix the mapping.
>>>>> I don't know the balloon driver, so I don't know why it was doing this,
>>>>> but what it was doing was Wrong and has been since 2014 with:
>>>>>
>>>>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>>>>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>>>>> Date: Thu Oct 9 15:29:27 2014 -0700
>>>>>
>>>>> mm/balloon_compaction: redesign ballooned pages management
>>>>>
>>>>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>>>>> them as ballooned pages using the mapcount field.
>>>>>
>>>> Asking myself why anybody would want to map balloon inflated pages into
>>>> user space (this just sounds plain wrong but my understanding to what
>>>> XEN balloon driver does might be limited), but I assume the easy fix
>>>> would be to revert
>>> I suspect the bug here is that the balloon driver is (ab)used for a
>>> second purpose
>>
>> Yes. And its name is alloc_xenballooned_pages().
>>
>
> Haven't had a look at the code yet, but would another temporary fix be
> to clear/set PG_offline when allocating/freeing a ballooned page?
> (assuming here that only such pages will be mapped to user space)
>
I guess something like this could do the trick if I understood it correctly:
diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c
index 39b229f9e256..d37dd5bb7a8f 100644
--- a/drivers/xen/balloon.c
+++ b/drivers/xen/balloon.c
@@ -604,6 +604,7 @@ int alloc_xenballooned_pages(int nr_pages, struct
page **pages)
while (pgno < nr_pages) {
page = balloon_retrieve(true);
if (page) {
+ __ClearPageOffline(page);
pages[pgno++] = page;
#ifdef CONFIG_XEN_HAVE_PVMMU
/*
@@ -645,8 +646,10 @@ void free_xenballooned_pages(int nr_pages, struct
page **pages)
mutex_lock(&balloon_mutex);
for (i = 0; i < nr_pages; i++) {
- if (pages[i])
+ if (pages[i]) {
+ __SetPageOffline(pages[i]);
balloon_append(pages[i]);
+ }
}
balloon_stats.target_unpopulated -= nr_pages;
At least this way, the pages allocated (and thus eventually mapped to
user space) would not be marked, but the other ones would remain marked
and could be excluded by makedumptool.
--
Thanks,
David / dhildenb
^ permalink raw reply related [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-12 19:46 ` David Hildenbrand
@ 2019-03-14 8:37 ` Juergen Gross
2019-03-14 14:12 ` Julien Grall
0 siblings, 1 reply; 16+ messages in thread
From: Juergen Gross @ 2019-03-14 8:37 UTC (permalink / raw)
To: David Hildenbrand, Boris Ostrovsky, Andrew Cooper,
Matthew Wilcox, Julien Grall
Cc: k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 12/03/2019 20:46, David Hildenbrand wrote:
> On 12.03.19 19:23, David Hildenbrand wrote:
>> On 12.03.19 19:02, Boris Ostrovsky wrote:
>>> On 3/12/19 1:24 PM, Andrew Cooper wrote:
>>>> On 12/03/2019 17:18, David Hildenbrand wrote:
>>>>> On 12.03.19 18:14, Matthew Wilcox wrote:
>>>>>> On Tue, Mar 12, 2019 at 05:05:39PM +0000, Julien Grall wrote:
>>>>>>> On 3/12/19 3:59 PM, Julien Grall wrote:
>>>>>>>> It looks like all the arm test for linus [1] and next [2] tree
>>>>>>>> are now failing. x86 seems to be mostly ok.
>>>>>>>>
>>>>>>>> The bisector fingered the following commit:
>>>>>>>>
>>>>>>>> commit 0ee930e6cafa048c1925893d0ca89918b2814f2c
>>>>>>>> Author: Matthew Wilcox <willy@infradead.org>
>>>>>>>> Date: Tue Mar 5 15:46:06 2019 -0800
>>>>>>>>
>>>>>>>> mm/memory.c: prevent mapping typed pages to userspace
>>>>>>>> Pages which use page_type must never be mapped to userspace as it would
>>>>>>>> destroy their page type. Add an explicit check for this instead of
>>>>>>>> assuming that kernel drivers always get this right.
>>>>>> Oh good, it found a real problem.
>>>>>>
>>>>>>> It turns out the problem is because the balloon driver will call
>>>>>>> __SetPageOffline() on allocated page. Therefore the page has a type and
>>>>>>> vm_insert_pages will deny the insertion.
>>>>>>>
>>>>>>> My knowledge is quite limited in this area. So I am not sure how we can
>>>>>>> solve the problem.
>>>>>>>
>>>>>>> I would appreciate if someone could provide input of to fix the mapping.
>>>>>> I don't know the balloon driver, so I don't know why it was doing this,
>>>>>> but what it was doing was Wrong and has been since 2014 with:
>>>>>>
>>>>>> commit d6d86c0a7f8ddc5b38cf089222cb1d9540762dc2
>>>>>> Author: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
>>>>>> Date: Thu Oct 9 15:29:27 2014 -0700
>>>>>>
>>>>>> mm/balloon_compaction: redesign ballooned pages management
>>>>>>
>>>>>> If ballooned pages are supposed to be mapped into userspace, you can't mark
>>>>>> them as ballooned pages using the mapcount field.
>>>>>>
>>>>> Asking myself why anybody would want to map balloon inflated pages into
>>>>> user space (this just sounds plain wrong but my understanding to what
>>>>> XEN balloon driver does might be limited), but I assume the easy fix
>>>>> would be to revert
>>>> I suspect the bug here is that the balloon driver is (ab)used for a
>>>> second purpose
>>>
>>> Yes. And its name is alloc_xenballooned_pages().
>>>
>>
>> Haven't had a look at the code yet, but would another temporary fix be
>> to clear/set PG_offline when allocating/freeing a ballooned page?
>> (assuming here that only such pages will be mapped to user space)
>>
>
> I guess something like this could do the trick if I understood it correctly:
>
> diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c
> index 39b229f9e256..d37dd5bb7a8f 100644
> --- a/drivers/xen/balloon.c
> +++ b/drivers/xen/balloon.c
> @@ -604,6 +604,7 @@ int alloc_xenballooned_pages(int nr_pages, struct
> page **pages)
> while (pgno < nr_pages) {
> page = balloon_retrieve(true);
> if (page) {
> + __ClearPageOffline(page);
> pages[pgno++] = page;
> #ifdef CONFIG_XEN_HAVE_PVMMU
> /*
> @@ -645,8 +646,10 @@ void free_xenballooned_pages(int nr_pages, struct
> page **pages)
> mutex_lock(&balloon_mutex);
>
> for (i = 0; i < nr_pages; i++) {
> - if (pages[i])
> + if (pages[i]) {
> + __SetPageOffline(pages[i]);
> balloon_append(pages[i]);
> + }
> }
>
> balloon_stats.target_unpopulated -= nr_pages;
>
>
> At least this way, the pages allocated (and thus eventually mapped to
> user space) would not be marked, but the other ones would remain marked
> and could be excluded by makedumptool.
>
I think this patch should do the trick. Julien, could you give it a
try? On x86 I can't reproduce your problem easily as dom0 is PV with
plenty of unpopulated pages for grant memory not suffering from
missing "offline" bit.
Juergen
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-14 8:37 ` Juergen Gross
@ 2019-03-14 14:12 ` Julien Grall
2019-03-14 14:14 ` David Hildenbrand
2019-03-14 14:15 ` Juergen Gross
0 siblings, 2 replies; 16+ messages in thread
From: Julien Grall @ 2019-03-14 14:12 UTC (permalink / raw)
To: Juergen Gross, David Hildenbrand, Boris Ostrovsky, Andrew Cooper,
Matthew Wilcox
Cc: k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
Hi,
On 3/14/19 8:37 AM, Juergen Gross wrote:
> On 12/03/2019 20:46, David Hildenbrand wrote:
>> On 12.03.19 19:23, David Hildenbrand wrote:
>>
>> I guess something like this could do the trick if I understood it correctly:
>>
>> diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c
>> index 39b229f9e256..d37dd5bb7a8f 100644
>> --- a/drivers/xen/balloon.c
>> +++ b/drivers/xen/balloon.c
>> @@ -604,6 +604,7 @@ int alloc_xenballooned_pages(int nr_pages, struct
>> page **pages)
>> while (pgno < nr_pages) {
>> page = balloon_retrieve(true);
>> if (page) {
>> + __ClearPageOffline(page);
>> pages[pgno++] = page;
>> #ifdef CONFIG_XEN_HAVE_PVMMU
>> /*
>> @@ -645,8 +646,10 @@ void free_xenballooned_pages(int nr_pages, struct
>> page **pages)
>> mutex_lock(&balloon_mutex);
>>
>> for (i = 0; i < nr_pages; i++) {
>> - if (pages[i])
>> + if (pages[i]) {
>> + __SetPageOffline(pages[i]);
>> balloon_append(pages[i]);
>> + }
>> }
>>
>> balloon_stats.target_unpopulated -= nr_pages;
>>
>>
>> At least this way, the pages allocated (and thus eventually mapped to
>> user space) would not be marked, but the other ones would remain marked
>> and could be excluded by makedumptool.
>>
>
> I think this patch should do the trick. Julien, could you give it a
> try? On x86 I can't reproduce your problem easily as dom0 is PV with
> plenty of unpopulated pages for grant memory not suffering from
> missing "offline" bit.
Sure. I managed to get the console working with the patch suggested by
David. Feel free to add my tested-by if when you resend it as is.
Cheers,
--
Julien Grall
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-14 14:12 ` Julien Grall
@ 2019-03-14 14:14 ` David Hildenbrand
2019-03-14 14:15 ` Juergen Gross
1 sibling, 0 replies; 16+ messages in thread
From: David Hildenbrand @ 2019-03-14 14:14 UTC (permalink / raw)
To: Julien Grall, Juergen Gross, Boris Ostrovsky, Andrew Cooper,
Matthew Wilcox
Cc: k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 14.03.19 15:12, Julien Grall wrote:
> Hi,
>
> On 3/14/19 8:37 AM, Juergen Gross wrote:
>> On 12/03/2019 20:46, David Hildenbrand wrote:
>>> On 12.03.19 19:23, David Hildenbrand wrote:
>>>
>>> I guess something like this could do the trick if I understood it correctly:
>>>
>>> diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c
>>> index 39b229f9e256..d37dd5bb7a8f 100644
>>> --- a/drivers/xen/balloon.c
>>> +++ b/drivers/xen/balloon.c
>>> @@ -604,6 +604,7 @@ int alloc_xenballooned_pages(int nr_pages, struct
>>> page **pages)
>>> while (pgno < nr_pages) {
>>> page = balloon_retrieve(true);
>>> if (page) {
>>> + __ClearPageOffline(page);
>>> pages[pgno++] = page;
>>> #ifdef CONFIG_XEN_HAVE_PVMMU
>>> /*
>>> @@ -645,8 +646,10 @@ void free_xenballooned_pages(int nr_pages, struct
>>> page **pages)
>>> mutex_lock(&balloon_mutex);
>>>
>>> for (i = 0; i < nr_pages; i++) {
>>> - if (pages[i])
>>> + if (pages[i]) {
>>> + __SetPageOffline(pages[i]);
>>> balloon_append(pages[i]);
>>> + }
>>> }
>>>
>>> balloon_stats.target_unpopulated -= nr_pages;
>>>
>>>
>>> At least this way, the pages allocated (and thus eventually mapped to
>>> user space) would not be marked, but the other ones would remain marked
>>> and could be excluded by makedumptool.
>>>
>>
>> I think this patch should do the trick. Julien, could you give it a
>> try? On x86 I can't reproduce your problem easily as dom0 is PV with
>> plenty of unpopulated pages for grant memory not suffering from
>> missing "offline" bit.
>
> Sure. I managed to get the console working with the patch suggested by
> David. Feel free to add my tested-by if when you resend it as is.
>
Thanks, I will send as proper patch later!
Cheers!
> Cheers,
>
--
Thanks,
David / dhildenb
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-14 14:12 ` Julien Grall
2019-03-14 14:14 ` David Hildenbrand
@ 2019-03-14 14:15 ` Juergen Gross
2019-03-14 14:16 ` David Hildenbrand
1 sibling, 1 reply; 16+ messages in thread
From: Juergen Gross @ 2019-03-14 14:15 UTC (permalink / raw)
To: Julien Grall, David Hildenbrand, Boris Ostrovsky, Andrew Cooper,
Matthew Wilcox
Cc: k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 14/03/2019 15:12, Julien Grall wrote:
> Hi,
>
> On 3/14/19 8:37 AM, Juergen Gross wrote:
>> On 12/03/2019 20:46, David Hildenbrand wrote:
>>> On 12.03.19 19:23, David Hildenbrand wrote:
>>>
>>> I guess something like this could do the trick if I understood it
>>> correctly:
>>>
>>> diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c
>>> index 39b229f9e256..d37dd5bb7a8f 100644
>>> --- a/drivers/xen/balloon.c
>>> +++ b/drivers/xen/balloon.c
>>> @@ -604,6 +604,7 @@ int alloc_xenballooned_pages(int nr_pages, struct
>>> page **pages)
>>> while (pgno < nr_pages) {
>>> page = balloon_retrieve(true);
>>> if (page) {
>>> + __ClearPageOffline(page);
>>> pages[pgno++] = page;
>>> #ifdef CONFIG_XEN_HAVE_PVMMU
>>> /*
>>> @@ -645,8 +646,10 @@ void free_xenballooned_pages(int nr_pages, struct
>>> page **pages)
>>> mutex_lock(&balloon_mutex);
>>>
>>> for (i = 0; i < nr_pages; i++) {
>>> - if (pages[i])
>>> + if (pages[i]) {
>>> + __SetPageOffline(pages[i]);
>>> balloon_append(pages[i]);
>>> + }
>>> }
>>>
>>> balloon_stats.target_unpopulated -= nr_pages;
>>>
>>>
>>> At least this way, the pages allocated (and thus eventually mapped to
>>> user space) would not be marked, but the other ones would remain marked
>>> and could be excluded by makedumptool.
>>>
>>
>> I think this patch should do the trick. Julien, could you give it a
>> try? On x86 I can't reproduce your problem easily as dom0 is PV with
>> plenty of unpopulated pages for grant memory not suffering from
>> missing "offline" bit.
>
> Sure. I managed to get the console working with the patch suggested by
> David. Feel free to add my tested-by if when you resend it as is.
David, could you please send a proper patch with your Sob?
Juergen
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS Re: [linux-linus bisection] complete test-arm64-arm64-xl-xsm)
2019-03-14 14:15 ` Juergen Gross
@ 2019-03-14 14:16 ` David Hildenbrand
0 siblings, 0 replies; 16+ messages in thread
From: David Hildenbrand @ 2019-03-14 14:16 UTC (permalink / raw)
To: Juergen Gross, Julien Grall, Boris Ostrovsky, Andrew Cooper,
Matthew Wilcox
Cc: k.khlebnikov, Stefano Stabellini, Kees Cook,
Konrad Rzeszutek Wilk, VMware, Inc.,
osstest service owner, linux-kernel, linux-mm, Julien Freche,
Nadav Amit, xen-devel
On 14.03.19 15:15, Juergen Gross wrote:
> On 14/03/2019 15:12, Julien Grall wrote:
>> Hi,
>>
>> On 3/14/19 8:37 AM, Juergen Gross wrote:
>>> On 12/03/2019 20:46, David Hildenbrand wrote:
>>>> On 12.03.19 19:23, David Hildenbrand wrote:
>>>>
>>>> I guess something like this could do the trick if I understood it
>>>> correctly:
>>>>
>>>> diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c
>>>> index 39b229f9e256..d37dd5bb7a8f 100644
>>>> --- a/drivers/xen/balloon.c
>>>> +++ b/drivers/xen/balloon.c
>>>> @@ -604,6 +604,7 @@ int alloc_xenballooned_pages(int nr_pages, struct
>>>> page **pages)
>>>> while (pgno < nr_pages) {
>>>> page = balloon_retrieve(true);
>>>> if (page) {
>>>> + __ClearPageOffline(page);
>>>> pages[pgno++] = page;
>>>> #ifdef CONFIG_XEN_HAVE_PVMMU
>>>> /*
>>>> @@ -645,8 +646,10 @@ void free_xenballooned_pages(int nr_pages, struct
>>>> page **pages)
>>>> mutex_lock(&balloon_mutex);
>>>>
>>>> for (i = 0; i < nr_pages; i++) {
>>>> - if (pages[i])
>>>> + if (pages[i]) {
>>>> + __SetPageOffline(pages[i]);
>>>> balloon_append(pages[i]);
>>>> + }
>>>> }
>>>>
>>>> balloon_stats.target_unpopulated -= nr_pages;
>>>>
>>>>
>>>> At least this way, the pages allocated (and thus eventually mapped to
>>>> user space) would not be marked, but the other ones would remain marked
>>>> and could be excluded by makedumptool.
>>>>
>>>
>>> I think this patch should do the trick. Julien, could you give it a
>>> try? On x86 I can't reproduce your problem easily as dom0 is PV with
>>> plenty of unpopulated pages for grant memory not suffering from
>>> missing "offline" bit.
>>
>> Sure. I managed to get the console working with the patch suggested by
>> David. Feel free to add my tested-by if when you resend it as is.
>
> David, could you please send a proper patch with your Sob?
>
Yes, on it :)
Cheers!
>
> Juergen
>
--
Thanks,
David / dhildenb
^ permalink raw reply [flat|nested] 16+ messages in thread
end of thread, other threads:[~2019-03-14 14:16 UTC | newest]
Thread overview: 16+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <E1h3Uiq-0002L6-Ij@osstest.test-lab.xenproject.org>
[not found] ` <80211e70-5f54-9421-8e8f-2a4fc758ce39@arm.com>
[not found] ` <46118631-61d4-adb6-6ffc-4e7c62ea3da9@arm.com>
2019-03-12 17:14 ` xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] [linux-linus bisection] complete test-arm64-arm64-xl-xsm) Matthew Wilcox
2019-03-12 17:18 ` David Hildenbrand
2019-03-12 17:24 ` [Xen-devel] xen: Can't insert balloon page into VM userspace (WAS " Andrew Cooper
2019-03-12 18:02 ` Boris Ostrovsky
2019-03-12 18:11 ` Andrew Cooper
2019-03-12 18:23 ` David Hildenbrand
2019-03-12 19:46 ` David Hildenbrand
2019-03-14 8:37 ` Juergen Gross
2019-03-14 14:12 ` Julien Grall
2019-03-14 14:14 ` David Hildenbrand
2019-03-14 14:15 ` Juergen Gross
2019-03-14 14:16 ` David Hildenbrand
2019-03-12 17:39 ` xen: Can't insert balloon page into VM userspace (WAS Re: [Xen-devel] " Julien Grall
2019-03-12 17:49 ` David Hildenbrand
2019-03-12 17:23 ` David Hildenbrand
2019-03-12 17:25 ` Nadav Amit
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).