| Message ID | 20240619141716.1785467-1-wsadowski@marvell.com |
|---|---|
| Headers | show |
| Series | Marvell HW overlay support for Cadence xSPI | expand |
On Wed, Jun 19, 2024 at 07:17:12AM -0700, Witold Sadowski wrote: > + while (t->len) { > + > + current_cycle_count = t->len > max_len ? max_len : t->len; cycle_count seems like a weird name? In general please try to avoid overusing the ternery operator, it doesn't help legibility. Just write normal conditional statements, or in this case use max(). > + cdns_xspi->current_xfer_qword += current_cycle_count / > + MRVL_XFER_QWORD_BYTECOUNT; > + if (current_cycle_count % MRVL_XFER_QWORD_BYTECOUNT) > + cdns_xspi->current_xfer_qword++; > + > + cdns_xspi->current_xfer_qword %= MRVL_XFER_QWORD_COUNT; > + } > + cs_change = t->cs_change; > + t->len -= current_cycle_count; > + } > + } This loop doesn't implement any of the delay stuff, ideally it either would or would reject transfers that request transfers.
On Fri, Jun 28, 2024 at 01:48:00PM +0000, Witold Sadowski wrote: > > This loop doesn't implement any of the delay stuff, ideally it either > > would or would reject transfers that request transfers. > I'm not sure about what kind of the delay You are referring to. Can You explain? The _delay fields in spi_transfer.
This patch series adds support for the second version of the Marvell hardware overlay for the Cadence xSPI IP block. The overlay is a hardware change made around the original xSPI block. It extends xSPI features with clock configuration, interrupt masking, and full-duplex, variable-length SPI operations. These functionalities allow the xSPI block to operate not only with memory devices but also with simple SPI devices and TPM devices. Example ACPI entry: Device (SPI0) { Name (_HID, "PRP0001") // ACPI_DT_NAMESPACE_HID Name (_UID, 0) Name (_DDN, "SPI controller 0") Name (_CCA, ONE) Method (_STA) {Return (0xF)} Name (_CRS, ResourceTemplate() { QWordMemory ( ResourceConsumer,// ResourceUsage PosDecode, // Decode MinFixed, // MinType MaxFixed, // MaxType NonCacheable, // MemType ReadWrite, // ReadWriteType 0, // AddressGranularity 0x804000000000, // MinAddress 0x804000001037, // MaxAddress 0, // AddressTranslation 0x1038) // AddressLength QWordMemory ( ResourceConsumer,// ResourceUsage PosDecode, // Decode MinFixed, // MinType MaxFixed, // MaxType NonCacheable, // MemType ReadWrite, // ReadWriteType 0, // AddressGranularity 0x804010000000, // MinAddress 0x804010000007, // MaxAddress 0, // AddressTranslation 0x8) // AddressLength QWordMemory ( ResourceConsumer,// ResourceUsage PosDecode, // Decode MinFixed, // MinType MaxFixed, // MaxType NonCacheable, // MemType ReadWrite, // ReadWriteType 0, // AddressGranularity 0x804000002000, // MinAddress 0x804000004027, // MaxAddress 0, // AddressTranslation 0x2028) // AddressLength QWordMemory ( ResourceConsumer,// ResourceUsage PosDecode, // Decode MinFixed, // MinType MaxFixed, // MaxType NonCacheable, // MemType ReadWrite, // ReadWriteType 0, // AddressGranularity 0x804000008000, // MinAddress 0x804000008237, // MaxAddress 0, // AddressTranslation 0x238) // AddressLength Interrupt(ResourceConsumer, Edge, ActiveHigh, Exclusive) { 0x7A } }) Name (_DSD, Package() { ToUUID("daffd814-6eba-4d8c-8a91-bc9bbf4aa301"), Package () { Package () { "compatible", "marvell,cn10-xspi-nor"}, Package () { "reg", 0x8040}, } }) } // SPI0 Changes: v9: Split into smaller patches: - Marvell overlay splitted into: PHY, Clock, Interrupt and SDMA ops - ACPI support splitted into resource mapping, CS parameter reading and tx/rx bus length Add separate ops and a few function pointers to distinguish between Cadence and Marvell: - SDMA read handler. - IRQ enable/disable handler - Separate mem_ops for Marvell xSPI Cleanup xfer code from magic numbers Add more descriptive commit msg for xfer block Use bitrev8 instead of custom bit reversal Rework Marvell SDMA read operations Add example ACPI entry v8: Rename xferbase to xfer Rework DLL reset, to return non inverted boolean value Rework STIG and SDMA status check, to return non inverted boolean value v7: Rebase patches to latest sources, changes in "Allow to read basic xSPI configuration from ACPI" Removed bugfix, as it was integrated to next tree from v6 v6: Fix item order in cdns,xspi.yaml v5: Rework cdns,xspi.yaml file Reword commit messages Move mamory mapping to ACPI patch Use devm_platform_ioremap_resource instead of two step mapping v4: Rename new Marvell registers to keep naming conventions Rename mrvl,xspi-nor to marvell,cnxx,xspi-nor Various fixed for cdns,xspi.yaml file: - Remove unnecesary parameters - Link register xferbase with marvell,cn10-xspi-nor - Move default values to .c file from device-tree Clock configuration optimization ACPI fixes: - Remove incorrect ACPI match table Added .data field to device_id, fixes for matching in ACPI and dtb case Minor style comment changes v3: Removed all kconfig changes Added device-tree mrvl,xspi-nor tag v2: Support for second overlay iteration v1: - v0: Initial support for v1 overlay Witold Sadowski (9): spi: dt-bindings: cadence: Add Marvell overlay bindings documentation for Cadence XSPI spi: cadence: Add static PHY configuration in Marvell overlay spi: cadence: Add clock configuration for Marvell xSPI overlay spi: cadence: Add Marvell SDMA operations spi: cadence: Add Marvell xSPI interrupt changes spi: cadence: Add Marvell xfer operation support spi: cadence: Change resource mapping spi: cadence: Change cs property reading. spi: cadence: Try to read spi-tx/rx-bus width property using ACPI .../devicetree/bindings/spi/cdns,xspi.yaml | 32 +- drivers/spi/spi-cadence-xspi.c | 665 +++++++++++++++++- 2 files changed, 677 insertions(+), 20 deletions(-)