From: Yong Wu <yong.wu@mediatek.com>
To: Robin Murphy <robin.murphy@arm.com>
Cc: youlin.pei@mediatek.com, anan.sun@mediatek.com,
Nicolas Boichat <drinkcat@chromium.org>,
cui.zhang@mediatek.com, srv_heupstream@mediatek.com,
chao.hao@mediatek.com, Joerg Roedel <joro@8bytes.org>,
edison.hsieh@mediatek.com, Will Deacon <will.deacon@arm.com>,
linux-kernel@vger.kernel.org, Evan Green <evgreen@chromium.org>,
Tomasz Figa <tfiga@google.com>,
iommu@lists.linux-foundation.org,
linux-mediatek@lists.infradead.org,
Matthias Brugger <matthias.bgg@gmail.com>,
linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH v3 3/7] iommu/mediatek: Use gather to achieve the tlb range flush
Date: Tue, 15 Oct 2019 13:26:06 +0800 [thread overview]
Message-ID: <1571117166.19130.83.camel@mhfsdcap03> (raw)
In-Reply-To: <f35c8a3a-0693-facf-2050-65d3f7628929@arm.com>
On Mon, 2019-10-14 at 15:21 +0100, Robin Murphy wrote:
> On 14/10/2019 07:38, Yong Wu wrote:
> > Use the iommu_gather mechanism to achieve the tlb range flush.
> > Gather the iova range in the "tlb_add_page", then flush the merged iova
> > range in iotlb_sync.
> >
> > Note: If iotlb_sync comes from iommu_iotlb_gather_add_page, we have to
> > avoid retry the lock since the spinlock have already been acquired.
>
> I think this could probably be even simpler - once the actual
> register-poking is all confined to mtk_iommu_tlb_sync(), you should be
> able get rid of the per-domain locking in map/unmap and just have a
> single per-IOMMU lock to serialise syncs. The io-pgtable code itself
> hasn't needed external locking for a while now.
This is more simpler! Thanks very much. I will try this.
The only concern is there is no lock in the iova_to_phys then, maybe use
the new lock instead.
>
> Robin.
>
> > Suggested-by: Tomasz Figa <tfiga@chromium.org>
> > Signed-off-by: Yong Wu <yong.wu@mediatek.com>
> > ---
> > 1) This is the special case backtrace:
> >
> > mtk_iommu_iotlb_sync+0x50/0xa0
> > mtk_iommu_tlb_flush_page_nosync+0x5c/0xd0
> > __arm_v7s_unmap+0x174/0x598
> > arm_v7s_unmap+0x30/0x48
> > mtk_iommu_unmap+0x50/0x78
> > __iommu_unmap+0xa4/0xf8
> >
> > 2) The checking "if (gather->start == ULONG_MAX) return;" also is
> > necessary. It will happened when unmap only go to _flush_walk, then
> > enter this tlb_sync.
> > ---
> > drivers/iommu/mtk_iommu.c | 29 +++++++++++++++++++++++++----
> > drivers/iommu/mtk_iommu.h | 1 +
> > 2 files changed, 26 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
> > index 5f594d6..8712afc 100644
> > --- a/drivers/iommu/mtk_iommu.c
> > +++ b/drivers/iommu/mtk_iommu.c
> > @@ -234,7 +234,12 @@ static void mtk_iommu_tlb_flush_page_nosync(struct iommu_iotlb_gather *gather,
> > unsigned long iova, size_t granule,
> > void *cookie)
> > {
> > - mtk_iommu_tlb_add_flush_nosync(iova, granule, granule, true, cookie);
> > + struct mtk_iommu_data *data = cookie;
> > + struct iommu_domain *domain = &data->m4u_dom->domain;
> > +
> > + data->is_in_tlb_gather_add_page = true;
> > + iommu_iotlb_gather_add_page(domain, gather, iova, granule);
> > + data->is_in_tlb_gather_add_page = false;
> > }
> >
> > static const struct iommu_flush_ops mtk_iommu_flush_ops = {
> > @@ -453,12 +458,28 @@ static void mtk_iommu_flush_iotlb_all(struct iommu_domain *domain)
> > static void mtk_iommu_iotlb_sync(struct iommu_domain *domain,
> > struct iommu_iotlb_gather *gather)
> > {
> > + struct mtk_iommu_data *data = mtk_iommu_get_m4u_data();
> > struct mtk_iommu_domain *dom = to_mtk_domain(domain);
> > + bool is_in_gather = data->is_in_tlb_gather_add_page;
> > + size_t length = gather->end - gather->start;
> > unsigned long flags;
> >
> > - spin_lock_irqsave(&dom->pgtlock, flags);
> > - mtk_iommu_tlb_sync(mtk_iommu_get_m4u_data());
> > - spin_unlock_irqrestore(&dom->pgtlock, flags);
> > + if (gather->start == ULONG_MAX)
> > + return;
> > +
> > + /*
> > + * Avoid acquire the lock when it's in gather_add_page since the lock
> > + * has already been held.
> > + */
> > + if (!is_in_gather)
> > + spin_lock_irqsave(&dom->pgtlock, flags);
> > +
> > + mtk_iommu_tlb_add_flush_nosync(gather->start, length, gather->pgsize,
> > + false, data);
> > + mtk_iommu_tlb_sync(data);
> > +
> > + if (!is_in_gather)
> > + spin_unlock_irqrestore(&dom->pgtlock, flags);
> > }
> >
> > static phys_addr_t mtk_iommu_iova_to_phys(struct iommu_domain *domain,
> > diff --git a/drivers/iommu/mtk_iommu.h b/drivers/iommu/mtk_iommu.h
> > index fc0f16e..d29af1d 100644
> > --- a/drivers/iommu/mtk_iommu.h
> > +++ b/drivers/iommu/mtk_iommu.h
> > @@ -58,6 +58,7 @@ struct mtk_iommu_data {
> > struct iommu_group *m4u_group;
> > bool enable_4GB;
> > bool tlb_flush_active;
> > + bool is_in_tlb_gather_add_page;
> >
> > struct iommu_device iommu;
> > const struct mtk_iommu_plat_data *plat_data;
> >
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
next prev parent reply other threads:[~2019-10-15 5:26 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-14 6:38 [PATCH v3 0/7] Improve tlb range flush Yong Wu
2019-10-14 6:38 ` [PATCH v3 1/7] iommu/mediatek: Correct the flush_iotlb_all callback Yong Wu
2019-10-14 14:06 ` Robin Murphy
2019-10-14 6:38 ` [PATCH v3 2/7] iommu/mediatek: Add pgtlock in the iotlb_sync Yong Wu
2019-10-14 6:38 ` [PATCH v3 3/7] iommu/mediatek: Use gather to achieve the tlb range flush Yong Wu
2019-10-14 14:21 ` Robin Murphy
2019-10-15 5:26 ` Yong Wu [this message]
2019-10-15 11:38 ` Robin Murphy
2019-10-14 6:38 ` [PATCH v3 4/7] iommu/mediatek: Delete the leaf in the tlb flush Yong Wu
2019-10-14 14:22 ` Robin Murphy
2019-10-15 5:25 ` Yong Wu
2019-10-15 11:24 ` Robin Murphy
2019-10-14 6:38 ` [PATCH v3 5/7] iommu/mediatek: Move the tlb_sync into tlb_flush Yong Wu
2019-10-14 6:38 ` [PATCH v3 6/7] iommu/mediatek: Use writel for TLB range invalidation Yong Wu
2019-10-14 14:04 ` Robin Murphy
2019-10-15 5:25 ` Yong Wu
2019-10-14 6:38 ` [PATCH v3 7/7] iommu/mediatek: Reduce the tlb flush timeout value Yong Wu
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=1571117166.19130.83.camel@mhfsdcap03 \
--to=yong.wu@mediatek.com \
--cc=anan.sun@mediatek.com \
--cc=chao.hao@mediatek.com \
--cc=cui.zhang@mediatek.com \
--cc=drinkcat@chromium.org \
--cc=edison.hsieh@mediatek.com \
--cc=evgreen@chromium.org \
--cc=iommu@lists.linux-foundation.org \
--cc=joro@8bytes.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mediatek@lists.infradead.org \
--cc=matthias.bgg@gmail.com \
--cc=robin.murphy@arm.com \
--cc=srv_heupstream@mediatek.com \
--cc=tfiga@google.com \
--cc=will.deacon@arm.com \
--cc=youlin.pei@mediatek.com \
/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
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox