From: Huacai Chen <chenhuacai@gmail.com>
To: Bjorn Helgaas <helgaas@kernel.org>
Cc: Huacai Chen <chenhuacai@loongson.cn>,
Bjorn Helgaas <bhelgaas@google.com>,
linux-pci <linux-pci@vger.kernel.org>,
Xuefeng Li <lixuefeng@loongson.cn>,
Jiaxun Yang <jiaxun.yang@flygoat.com>
Subject: Re: [PATCH V3 3/4] PCI: Improve the MRRS quirk for LS7A
Date: Sun, 27 Jun 2021 18:25:04 +0800 [thread overview]
Message-ID: <CAAhV-H68tBbTxoy-qOP4F3KDWEjunZMC-v3dAPWfU==NCMBbyA@mail.gmail.com> (raw)
In-Reply-To: <20210625222204.GA3657225@bjorn-Precision-5520>
Hi, Bjorn,
On Sat, Jun 26, 2021 at 6:22 AM Bjorn Helgaas <helgaas@kernel.org> wrote:
>
> On Fri, Jun 25, 2021 at 05:30:29PM +0800, Huacai Chen wrote:
> > In new revision of LS7A, some PCIe ports support larger value than 256,
> > but their maximum supported MRRS values are not detectable. Moreover,
> > the current loongson_mrrs_quirk() cannot avoid devices increasing its
> > MRRS after pci_enable_device(), and some devices (e.g. Realtek 8169)
> > will actually set a big value in its driver. So the only possible way is
> > configure MRRS of all devices in BIOS, and add a PCI device flag (i.e.,
> > PCI_DEV_FLAGS_NO_INCREASE_MRRS) to stop the increasing MRRS operations.
> >
> > However, according to PCIe Spec, it is legal for an OS to program any
> > value for MRRS, and it is also legal for an endpoint to generate a Read
> > Request with any size up to its MRRS. As the hardware engineers says,
> > the root cause here is LS7A doesn't break up large read requests (Yes,
> > that is a problem in the LS7A design).
>
> "LS7A doesn't break up large read requests" claims to be a root cause,
> but you haven't yet said what the actual *problem* is.
>
> Is the problem that an endpoint reports a malformed TLP because it
> received a completion bigger than it can handle? Is it that the LS7A
> root port reports some kind of error if it receives a Memory Read
> request with a size that's "too big"? Maybe the LS7A doesn't know
> what to do when it receives a Memory Read request with MRRS > MPS?
> What exactly happens when the problem occurs?
The hardware engineer said that the problem is: LS7A PCIe port reports
CA (Completer Abort) if it receives a Memory Read
request with a size that's "too big".
Huacai
>
> MRRS applies only to the read request. It is not directly related to
> the size of the completions that carry the data back to the device
> (except that obviously you shouldn't get a completion larger than the
> read you requested).
>
> The setting that directly controls the size of completions is MPS
> (Max_Payload_Size). One reason to break up read requests is because
> the endpoint's buffers can't accommodate big TLPs. One way to deal
> with that is to set MPS in the hierarchy to a smaller value. Then the
> root port must ensure that no TLP exceeds the MPS size, regardless of
> what the MRRS in the read request was.
>
> For example, if the endpoint's MRRS=4096 and the hierarchy's MPS=128,
> it's up to the root port to break up completions into 128-byte chunks.
>
> It's also possible to set the endpoint's MRRS=128, which means reads
> to main memory will never receive completions larger than 128 bytes.
> But it does NOT guarantee that a peer-to-peer DMA from another device
> will be limited to 128 bytes. The other device is allowed to generate
> Memory Write TLPs with payloads up to its MPS size, and MRRS is not
> involved at all.
>
> It's not clear yet whether the LS7A problem is with MRRS, with MPS, or
> with some combination. It's important to understand exactly what is
> broken here so the quirk doesn't get in the way of future changes to
> the generic MRRS and MPS configuration.
>
> Here's a good overview:
>
> https://www.xilinx.com/support/documentation/white_papers/wp350.pdf
>
> > Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
> > ---
> > drivers/pci/pci.c | 5 +++++
> > drivers/pci/quirks.c | 8 +++++++-
> > include/linux/pci.h | 2 ++
> > 3 files changed, 14 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > index b717680377a9..6f0d2f5b6f30 100644
> > --- a/drivers/pci/pci.c
> > +++ b/drivers/pci/pci.c
> > @@ -5802,6 +5802,11 @@ int pcie_set_readrq(struct pci_dev *dev, int rq)
> >
> > v = (ffs(rq) - 8) << 12;
> >
> > + if (dev->dev_flags & PCI_DEV_FLAGS_NO_INCREASE_MRRS) {
> > + if (rq > pcie_get_readrq(dev))
> > + return -EINVAL;
> > + }
> > +
> > ret = pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
> > PCI_EXP_DEVCTL_READRQ, v);
> >
> > diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> > index dee4798a49fc..8284480dc7e4 100644
> > --- a/drivers/pci/quirks.c
> > +++ b/drivers/pci/quirks.c
> > @@ -263,7 +263,13 @@ static void loongson_mrrs_quirk(struct pci_dev *dev)
> > * anything larger than this. So force this limit on
> > * any devices attached under these ports.
> > */
> > - if (pci_match_id(bridge_devids, bridge)) {
> > + if (bridge && pci_match_id(bridge_devids, bridge)) {
> > + dev->dev_flags |= PCI_DEV_FLAGS_NO_INCREASE_MRRS;
> > +
> > + if (pcie_bus_config == PCIE_BUS_DEFAULT ||
> > + pcie_bus_config == PCIE_BUS_TUNE_OFF)
> > + break;
> > +
> > if (pcie_get_readrq(dev) > 256) {
> > pci_info(dev, "limiting MRRS to 256\n");
> > pcie_set_readrq(dev, 256);
> > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > index 24306504226a..5e0ec3e4318b 100644
> > --- a/include/linux/pci.h
> > +++ b/include/linux/pci.h
> > @@ -227,6 +227,8 @@ enum pci_dev_flags {
> > PCI_DEV_FLAGS_NO_FLR_RESET = (__force pci_dev_flags_t) (1 << 10),
> > /* Don't use Relaxed Ordering for TLPs directed at this device */
> > PCI_DEV_FLAGS_NO_RELAXED_ORDERING = (__force pci_dev_flags_t) (1 << 11),
> > + /* Don't increase BIOS's MRRS configuration */
> > + PCI_DEV_FLAGS_NO_INCREASE_MRRS = (__force pci_dev_flags_t) (1 << 12),
> > };
> >
> > enum pci_irq_reroute_variant {
> > --
> > 2.27.0
> >
next prev parent reply other threads:[~2021-06-27 10:25 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-25 9:30 [PATCH V3 0/4] PCI: Loongson-related pci quirks Huacai Chen
2021-06-25 9:30 ` [PATCH V3 1/4] PCI/portdrv: Don't disable device during shutdown Huacai Chen
2021-06-25 20:45 ` Bjorn Helgaas
2021-06-28 9:32 ` Huacai Chen
2021-06-25 9:30 ` [PATCH V3 2/4] PCI: Move loongson pci quirks to quirks.c Huacai Chen
2021-06-26 15:39 ` Bjorn Helgaas
2021-06-27 9:54 ` Huacai Chen
2021-06-25 9:30 ` [PATCH V3 3/4] PCI: Improve the MRRS quirk for LS7A Huacai Chen
2021-06-25 22:22 ` Bjorn Helgaas
2021-06-26 15:48 ` Bjorn Helgaas
2021-06-27 10:27 ` Huacai Chen
2021-06-27 10:25 ` Huacai Chen [this message]
2021-06-28 20:51 ` Bjorn Helgaas
2021-06-29 2:00 ` Huacai Chen
2021-06-29 2:12 ` Bjorn Helgaas
2021-06-29 3:32 ` Huacai Chen
2021-06-29 3:38 ` Jiaxun Yang
2021-06-25 9:30 ` [PATCH V3 4/4] PCI: Add quirk for multifunction devices of LS7A Huacai Chen
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='CAAhV-H68tBbTxoy-qOP4F3KDWEjunZMC-v3dAPWfU==NCMBbyA@mail.gmail.com' \
--to=chenhuacai@gmail.com \
--cc=bhelgaas@google.com \
--cc=chenhuacai@loongson.cn \
--cc=helgaas@kernel.org \
--cc=jiaxun.yang@flygoat.com \
--cc=linux-pci@vger.kernel.org \
--cc=lixuefeng@loongson.cn \
/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;
as well as URLs for NNTP newsgroup(s).