From: Rob Herring <robh@kernel.org>
To: Kishon Vijay Abraham I <kishon@ti.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>,
Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
Arnd Bergmann <arnd@arndb.de>, Tom Joseph <tjoseph@cadence.com>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
PCI <linux-pci@vger.kernel.org>,
devicetree@vger.kernel.org,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
linux-omap <linux-omap@vger.kernel.org>,
"moderated list:ARM/FREESCALE IMX / MXC ARM ARCHITECTURE"
<linux-arm-kernel@lists.infradead.org>
Subject: Re: [PATCH v4 03/14] PCI: cadence: Add support to use custom read and write accessors
Date: Thu, 21 May 2020 16:17:09 -0600 [thread overview]
Message-ID: <CAL_JsqK8xRTxuJ39yPMzE1JMBrYccXB5QqsSvn7HosvrEQGdWQ@mail.gmail.com> (raw)
In-Reply-To: <37d2d6c3-232d-adb8-4e0b-e0c699ec435a@ti.com>
On Thu, May 21, 2020 at 7:33 AM Kishon Vijay Abraham I <kishon@ti.com> wrote:
>
> Hi Rob,
>
> On 5/21/2020 3:37 AM, Rob Herring wrote:
> > On Wed, May 06, 2020 at 08:44:18PM +0530, Kishon Vijay Abraham I wrote:
> >> Add support to use custom read and write accessors. Platforms that
> >> don't support half word or byte access or any other constraint
> >> while accessing registers can use this feature to populate custom
> >> read and write accessors. These custom accessors are used for both
> >> standard register access and configuration space register access of
> >> the PCIe host bridge.
> >>
> >> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
> >> ---
> >> drivers/pci/controller/cadence/pcie-cadence.h | 107 +++++++++++++++---
> >> 1 file changed, 94 insertions(+), 13 deletions(-)
> >
> > Actually, take back my R-by...
> >
> >>
> >> diff --git a/drivers/pci/controller/cadence/pcie-cadence.h b/drivers/pci/controller/cadence/pcie-cadence.h
> >> index df14ad002fe9..70b6b25153e8 100644
> >> --- a/drivers/pci/controller/cadence/pcie-cadence.h
> >> +++ b/drivers/pci/controller/cadence/pcie-cadence.h
> >> @@ -223,6 +223,11 @@ enum cdns_pcie_msg_routing {
> >> MSG_ROUTING_GATHER,
> >> };
> >>
> >> +struct cdns_pcie_ops {
> >> + u32 (*read)(void __iomem *addr, int size);
> >> + void (*write)(void __iomem *addr, int size, u32 value);
> >> +};
> >> +
> >> /**
> >> * struct cdns_pcie - private data for Cadence PCIe controller drivers
> >> * @reg_base: IO mapped register base
> >> @@ -239,7 +244,7 @@ struct cdns_pcie {
> >> int phy_count;
> >> struct phy **phy;
> >> struct device_link **link;
> >> - const struct cdns_pcie_common_ops *ops;
> >> + const struct cdns_pcie_ops *ops;
> >> };
> >>
> >> /**
> >> @@ -299,69 +304,145 @@ struct cdns_pcie_ep {
> >> /* Register access */
> >> static inline void cdns_pcie_writeb(struct cdns_pcie *pcie, u32 reg, u8 value)
> >> {
> >> - writeb(value, pcie->reg_base + reg);
> >> + void __iomem *addr = pcie->reg_base + reg;
> >> +
> >> + if (pcie->ops && pcie->ops->write) {
> >> + pcie->ops->write(addr, 0x1, value);
> >> + return;
> >> + }
> >> +
> >> + writeb(value, addr);
> >> }
> >>
> >> static inline void cdns_pcie_writew(struct cdns_pcie *pcie, u32 reg, u16 value)
> >> {
> >> - writew(value, pcie->reg_base + reg);
> >> + void __iomem *addr = pcie->reg_base + reg;
> >> +
> >> + if (pcie->ops && pcie->ops->write) {
> >> + pcie->ops->write(addr, 0x2, value);
> >> + return;
> >> + }
> >> +
> >> + writew(value, addr);
> >> }
> >
> > cdns_pcie_writeb and cdns_pcie_writew are used, so remove them.
> >
> >>
> >> static inline void cdns_pcie_writel(struct cdns_pcie *pcie, u32 reg, u32 value)
> >> {
> >> - writel(value, pcie->reg_base + reg);
> >> + void __iomem *addr = pcie->reg_base + reg;
> >> +
> >> + if (pcie->ops && pcie->ops->write) {
> >> + pcie->ops->write(addr, 0x4, value);
> >> + return;
> >> + }
> >> +
> >> + writel(value, addr);
> >
> > writel isn't broken for you, so you don't need this either.
> >
> >> }
> >>
> >> static inline u32 cdns_pcie_readl(struct cdns_pcie *pcie, u32 reg)
> >> {
> >> - return readl(pcie->reg_base + reg);
> >> + void __iomem *addr = pcie->reg_base + reg;
> >> +
> >> + if (pcie->ops && pcie->ops->read)
> >> + return pcie->ops->read(addr, 0x4);
> >> +
> >> + return readl(addr);
> >
> > And neither is readl.
>
> Sure, I'll remove all the unused functions and avoid using ops for readl and
> writel.
> >
> >> }
> >>
> >> /* Root Port register access */
> >> static inline void cdns_pcie_rp_writeb(struct cdns_pcie *pcie,
> >> u32 reg, u8 value)
> >> {
> >> - writeb(value, pcie->reg_base + CDNS_PCIE_RP_BASE + reg);
> >> + void __iomem *addr = pcie->reg_base + CDNS_PCIE_RP_BASE + reg;
> >> +
> >> + if (pcie->ops && pcie->ops->write) {
> >> + pcie->ops->write(addr, 0x1, value);
> >> + return;
> >> + }
> >> +
> >> + writeb(value, addr);
> >> }
> >>
> >> static inline void cdns_pcie_rp_writew(struct cdns_pcie *pcie,
> >> u32 reg, u16 value)
> >> {
> >> - writew(value, pcie->reg_base + CDNS_PCIE_RP_BASE + reg);
> >> + void __iomem *addr = pcie->reg_base + CDNS_PCIE_RP_BASE + reg;
> >> +
> >> + if (pcie->ops && pcie->ops->write) {
> >> + pcie->ops->write(addr, 0x2, value);
> >> + return;
> >> + }
> >> +
> >> + writew(value, addr);
> >
> > You removed 2 out of 3 calls to this. I think I'd just make the root
> > port writes always be 32-bit. It is all just one time init stuff
> > anyways.
> >
> > Either rework the calls to assemble the data into 32-bits or keep these
> > functions and do the RMW here.
>
> The problem with assembling data into 32-bits is we have to read/write with
> different offsets. We'll give PCI_VENDOR_ID offset for modifying deviceID,
> PCI_INTERRUPT_LINE for modifying INTERRUPT_PIN which might get non-intuitive.
> Similarly in endpoint we read and write to MSI_FLAGS (which is at offset 2) we
> have to directly use MSI capability offset.
>
> And doing RMW in the accessors would mean the same RMW op is repeated. So if we
> just have cdns_pcie_rp_writeb() and cdns_pcie_rp_writew(), the same code will
> be repeated here twice.
Why repeated? Just copy what the config accessors do:
static inline void cdns_pcie_write_sz(u32 val, void __iomem *addr, int size)
{
u32 tmp, mask, where = (unsigned long)addr & 0x3;
addr -= where;
mask = ~(((1 << (size * 8)) - 1) << (where * 8));
tmp = readl(addr) & mask;
tmp |= val << (where * 8);
writel(tmp, addr);
}
/* Root Port register access */
static inline void cdns_pcie_rp_writeb(struct cdns_pcie *pcie,
u32 reg, u8 value)
{
cdns_pcie_write_sz(value, pcie->reg_base + CDNS_PCIE_RP_BASE + reg, 1);
}
static inline void cdns_pcie_rp_writew(struct cdns_pcie *pcie,
u32 reg, u16 value)
{
cdns_pcie_write_sz(value, pcie->reg_base + CDNS_PCIE_RP_BASE + reg, 2);
}
>
> IMHO using ops is a lot cleaner for these cases. IMHO except for removing
> unused functions and not changing readl/writel, others should use ops.
Trying to read the DW PCI driver I don't agree...
Again, unless doing RMW is fundamentally broken (like it is for config
space at runtime), then you only want to do it on broken h/w and ops
would be appropriate. It is all mostly one time setup, so doing a few
extra reads isn't a big deal. If you really care about speed on that,
probably should use the _relaxed accessor variants.
I'm being hopeful the Cadence IP doesn't become the mess that DW is.
Rob
next prev parent reply other threads:[~2020-05-21 22:17 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-06 15:14 [PATCH v4 00/14] Add PCIe support to TI's J721E SoC Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 01/14] PCI: cadence: Fix cdns_pcie_{host|ep}_setup() error path Kishon Vijay Abraham I
2020-05-20 20:59 ` Rob Herring
2020-05-06 15:14 ` [PATCH v4 02/14] linux/kernel.h: Add PTR_ALIGN_DOWN macro Kishon Vijay Abraham I
2020-05-20 21:00 ` Rob Herring
2020-05-06 15:14 ` [PATCH v4 03/14] PCI: cadence: Add support to use custom read and write accessors Kishon Vijay Abraham I
2020-05-20 21:02 ` Rob Herring
2020-05-20 22:07 ` Rob Herring
2020-05-21 13:33 ` Kishon Vijay Abraham I
2020-05-21 22:17 ` Rob Herring [this message]
2020-05-22 3:36 ` Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 04/14] PCI: cadence: Add support to start link and verify link status Kishon Vijay Abraham I
2020-05-20 21:06 ` Rob Herring
2020-05-06 15:14 ` [PATCH v4 05/14] PCI: cadence: Add read/write accessors to perform only 32-bit accesses Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 06/14] PCI: cadence: Allow pci_host_bridge to have custom pci_ops Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 07/14] PCI: cadence: Add new *ops* for CPU addr fixup Kishon Vijay Abraham I
2020-05-20 21:34 ` Rob Herring
2020-05-21 11:34 ` Kishon Vijay Abraham I
2020-05-22 16:45 ` Rob Herring
2020-05-23 1:24 ` Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 08/14] PCI: cadence: Fix updating Vendor ID and Subsystem Vendor ID register Kishon Vijay Abraham I
2020-05-20 21:36 ` Rob Herring
2020-05-06 15:14 ` [PATCH v4 09/14] PCI: cadence: Add MSI-X support to Endpoint driver Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 10/14] dt-bindings: PCI: Add host mode dt-bindings for TI's J721E SoC Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 11/14] dt-bindings: PCI: Add EP " Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 12/14] PCI: j721e: Add TI J721E PCIe driver Kishon Vijay Abraham I
2020-05-06 15:14 ` [PATCH v4 13/14] misc: pci_endpoint_test: Add J721E in pci_device_id table Kishon Vijay Abraham I
2020-05-20 22:12 ` Rob Herring
2020-05-06 15:14 ` [PATCH v4 14/14] MAINTAINERS: Add Kishon Vijay Abraham I for TI J721E SoC PCIe Kishon Vijay Abraham I
2020-05-20 22:12 ` Rob Herring
2020-05-18 11:14 ` [PATCH v4 00/14] Add PCIe support to TI's J721E SoC Kishon Vijay Abraham I
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=CAL_JsqK8xRTxuJ39yPMzE1JMBrYccXB5QqsSvn7HosvrEQGdWQ@mail.gmail.com \
--to=robh@kernel.org \
--cc=arnd@arndb.de \
--cc=bhelgaas@google.com \
--cc=devicetree@vger.kernel.org \
--cc=gregkh@linuxfoundation.org \
--cc=kishon@ti.com \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-omap@vger.kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=lorenzo.pieralisi@arm.com \
--cc=tjoseph@cadence.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;
as well as URLs for NNTP newsgroup(s).