> On May 3, 2020, at 2:39 PM, Joerg Roedel wrote:
>
> Can I add your Tested-by when I
> send them to the mailing list tomorrow?
Sure. Feel free to add,
Tested-by: Qian Cai
___
iommu mailing list
iommu@lists.linux-foundation.org
Hi Qian,
On Sun, May 03, 2020 at 09:04:03AM -0400, Qian Cai wrote:
> > On Apr 29, 2020, at 7:20 AM, Joerg Roedel wrote:
> > Can you please test this branch:
> >
> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/joro/linux.git/log/?h=amd-iommu-fixes
> >
> > It has the previous fix in
> On Apr 29, 2020, at 7:20 AM, Joerg Roedel wrote:
>
> On Mon, Apr 20, 2020 at 09:26:12AM -0400, Qian Cai wrote:
>> No dice. There could be some other races. For example,
>
> Can you please test this branch:
>
>
>
> On Apr 29, 2020, at 7:20 AM, Joerg Roedel wrote:
>
> On Mon, Apr 20, 2020 at 09:26:12AM -0400, Qian Cai wrote:
>> No dice. There could be some other races. For example,
>
> Can you please test this branch:
>
>
>
On Mon, Apr 20, 2020 at 09:26:12AM -0400, Qian Cai wrote:
> No dice. There could be some other races. For example,
Can you please test this branch:
https://git.kernel.org/pub/scm/linux/kernel/git/joro/linux.git/log/?h=amd-iommu-fixes
It has the previous fix in it and a couple more to
Hi Qian,
On Mon, Apr 20, 2020 at 09:26:12AM -0400, Qian Cai wrote:
>
> No dice. There could be some other races. For example,
Okay, I think I know what is happening. The increase_address_space()
function increases the address space, but does not update the
DTE and does not flush the old DTE
> On Apr 18, 2020, at 2:34 PM, Joerg Roedel wrote:
>
> On Sat, Apr 18, 2020 at 09:01:35AM -0400, Qian Cai wrote:
>> Hard to tell without testing further. I’ll leave that optimization in
>> the future, and focus on fixing those races first.
>
> Yeah right, we should fix the existing races
> On Apr 18, 2020, at 2:34 PM, Joerg Roedel wrote:
>
> On Sat, Apr 18, 2020 at 09:01:35AM -0400, Qian Cai wrote:
>> Hard to tell without testing further. I’ll leave that optimization in
>> the future, and focus on fixing those races first.
>
> Yeah right, we should fix the existing races
On Sat, Apr 18, 2020 at 09:01:35AM -0400, Qian Cai wrote:
> Hard to tell without testing further. I’ll leave that optimization in
> the future, and focus on fixing those races first.
Yeah right, we should fix the existing races first before introducing
new ones ;)
Btw, THANKS A LOT for tracking
> On Apr 18, 2020, at 8:10 AM, Joerg Roedel wrote:
>
> Yes, your patch still looks racy. You need to atomically read
> domain->pt_root to a stack variable and derive the pt_root pointer and
> the mode from that variable instead of domain->pt_root directly. If you
> read the domain->pt_root
On Thu, Apr 16, 2020 at 09:42:41PM -0400, Qian Cai wrote:
> So, this is still not enough that would still trigger storage driver offline
> under
> memory pressure for a bit longer. It looks to me that in fetch_pte() there are
> could still racy?
Yes, your patch still looks racy. You need to
> On Apr 13, 2020, at 9:36 PM, Qian Cai wrote:
>
>
>
>> On Apr 8, 2020, at 10:19 AM, Joerg Roedel wrote:
>>
>> Hi Qian,
>>
>> On Tue, Apr 07, 2020 at 11:36:05AM -0400, Qian Cai wrote:
>>> After further testing, the change along is insufficient. What I am chasing
>>> right
>>> now is the
> On Apr 8, 2020, at 10:19 AM, Joerg Roedel wrote:
>
> Hi Qian,
>
> On Tue, Apr 07, 2020 at 11:36:05AM -0400, Qian Cai wrote:
>> After further testing, the change along is insufficient. What I am chasing
>> right
>> now is the swap device will go offline after heavy memory pressure below.
Hi Qian,
On Tue, Apr 07, 2020 at 11:36:05AM -0400, Qian Cai wrote:
> After further testing, the change along is insufficient. What I am chasing
> right
> now is the swap device will go offline after heavy memory pressure below. The
> symptom is similar to what we have in the commit,
>
>
> On Apr 6, 2020, at 10:12 PM, Qian Cai wrote:
>
> fetch_pte() could race with increase_address_space() because it held no
> lock from iommu_unmap_page(). On the CPU that runs fetch_pte() it could
> see a stale domain->pt_root and a new increased domain->mode from
> increase_address_space().
fetch_pte() could race with increase_address_space() because it held no
lock from iommu_unmap_page(). On the CPU that runs fetch_pte() it could
see a stale domain->pt_root and a new increased domain->mode from
increase_address_space(). As the result, it could trigger invalid
accesses later on. Fix
16 matches
Mail list logo