URL
https://opencores.org/ocsvn/openrisc/openrisc/trunk
Subversion Repositories openrisc
[/] [openrisc/] [trunk/] [rtos/] [ecos-2.0/] [packages/] [devs/] [usb/] [nec_upd985xx/] [v2_0/] [src/] [usbs_upd985xx.c] - Rev 454
Go to most recent revision | Compare with Previous | Blame | View Log
//========================================================================== // // usbs_upd985xx.c // // Driver for the NEC uPD985xx USB device // //========================================================================== //####ECOSGPLCOPYRIGHTBEGIN#### // ------------------------------------------- // This file is part of eCos, the Embedded Configurable Operating System. // Copyright (C) 2002 Bart Veer // Copyright (C) 1998, 1999, 2000, 2001, 2002 Red Hat, Inc. // // eCos is free software; you can redistribute it and/or modify it under // the terms of the GNU General Public License as published by the Free // Software Foundation; either version 2 or (at your option) any later version. // // eCos is distributed in the hope that it will be useful, but WITHOUT ANY // WARRANTY; without even the implied warranty of MERCHANTABILITY or // FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License // for more details. // // You should have received a copy of the GNU General Public License along // with eCos; if not, write to the Free Software Foundation, Inc., // 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA. // // As a special exception, if other files instantiate templates or use macros // or inline functions from this file, or you compile this file and link it // with other works to produce a work based on this file, this file does not // by itself cause the resulting work to be covered by the GNU General Public // License. However the source code for this file must still be made available // in accordance with section (3) of the GNU General Public License. // // This exception does not invalidate any other reasons why a work based on // this file might be covered by the GNU General Public License. // // Alternative licenses for eCos may be arranged by contacting Red Hat, Inc. // at http://sources.redhat.com/ecos/ecos-license/ // ------------------------------------------- //####ECOSGPLCOPYRIGHTEND#### //========================================================================== //#####DESCRIPTIONBEGIN#### // // Author(s): bartv // Contributors: bartv // Date: 2001-05-22 // // This code implements support for the on-chip USB port on the NEC // uPD985xx family of processors. The code has been developed on the // uPD98503 and may or may not work on other members of the uPD985xx // family. // //####DESCRIPTIONEND#### //========================================================================== #include <cyg/infra/cyg_type.h> #include <cyg/infra/cyg_ass.h> #include <cyg/infra/cyg_trac.h> #include <cyg/infra/diag.h> #include <pkgconf/hal_mips_upd985xx.h> #include <pkgconf/devs_usb_upd985xx.h> #include <cyg/hal/drv_api.h> #include <cyg/hal/hal_arch.h> #include <cyg/hal/hal_io.h> #include <cyg/hal/hal_cache.h> #include <cyg/error/codes.h> #include <cyg/io/usb/usb.h> #include <cyg/io/usb/usbs.h> // For memcpy() #include <string.h> // ---------------------------------------------------------------------------- // Toplevel FIXME's. // // The device supports remote wakeups, but this driver does not. Note that // the device GET_STATUS, SET_FEATURE and CLEAR_FEATURE operations are // affected by remote wakeup support. // ---------------------------------------------------------------------------- // Debugging-related odds and ends. #if 0 # define DBG(a) diag_printf a #else # define DBG(a) #endif // ---------------------------------------------------------------------------- // Hardware definitions. // // The NEC uPD985Xx on-chip USB device provides the following: // // endpoint 0 - control messages only // endpoint 1 - isochronous transmits // endpoint 2 - isochronous receives // endpoint 3 - bulk transmits // endpoint 4 - bulk receives // endpoint 5 - interrupt transmits // endpoint 6 - interrupt receives // All acess to the USB controller registers goes via the IBUS, which // always runs little-endian. Hence when the CPU is running // little-endian no extra work is needed, but when running big-endian // all register updates involve swapping. #ifdef CYGPKG_HAL_MIPS_LSBFIRST # define IBUS_SWAP32(_a_) (_a_) # define IBUS_SWAPPTR(_type_, _a_) (_a_) #else # error IBUS_SWAP32() needs to be defined and tested #endif // Move an address to kseg1. Or'ing in the relevant bits means // that this macro will work even if the specified address is // already in kseg1 #define MIPS_TO_UNCACHED(_a_) ((void*)(((cyg_uint32)(_a_)) | MIPS_KSEG1_BASE)) // For now access the various registers directly. A structure might // be marginally more inefficient in that if a function accesses // several registers this could be handled by a single base plus // offsets, rather than by separate addresses. #define USBS_REGISTER(_a_) ((volatile cyg_uint32*)(MIPS_IO_BASE + UPD985XX_SYSUSB_OFF + (_a_))) #define USBS_ADDRREG(_type_, _a_) ((_type_* volatile*)(MIPS_IO_BASE + UPD985XX_SYSUSB_OFF + (_a_))) #define USBS_GMR USBS_REGISTER(0x0000) #define USBS_VER USBS_REGISTER(0x0004) #define USBS_GSR1 USBS_REGISTER(0x0010) #define USBS_IMR1 USBS_REGISTER(0x0014) #define USBS_GSR2 USBS_REGISTER(0x0018) #define USBS_IMR2 USBS_REGISTER(0x001C) #define EP0_CR USBS_REGISTER(0x0020) #define EP1_CR USBS_REGISTER(0x0024) #define EP2_CR USBS_REGISTER(0x0028) #define EP3_CR USBS_REGISTER(0x002C) #define EP4_CR USBS_REGISTER(0x0030) #define EP5_CR USBS_REGISTER(0x0034) #define EP6_CR USBS_REGISTER(0x0038) #define USBS_CMR USBS_REGISTER(0x0040) #define USBS_CA USBS_ADDRREG(void, 0x0044) #define USBS_TEPSR USBS_REGISTER(0x0048) #define USBS_RP0IR USBS_REGISTER(0x0050) #define USBS_RP0AR USBS_ADDRREG(RxBufferDescriptor, 0x0054) #define USBS_RP1IR USBS_REGISTER(0x0058) #define USBS_RP1AR USBS_ADDRREG(RxBufferDescriptor, 0x005C) #define USBS_RP2IR USBS_REGISTER(0x0060) #define USBS_RP2AR USBS_ADDRREG(RxBufferDescriptor, 0x0064) #define USBS_TMSA USBS_ADDRREG(TxMailbox, 0x0070) #define USBS_TMBA USBS_ADDRREG(TxMailbox, 0x0074) #define USBS_TMRA USBS_ADDRREG(TxMailbox, 0x0078) #define USBS_TMWA USBS_ADDRREG(TxMailbox, 0x007C) #define USBS_RMSA USBS_ADDRREG(RxMailbox, 0x0080) #define USBS_RMBA USBS_ADDRREG(RxMailbox, 0x0084) #define USBS_RMRA USBS_ADDRREG(RxMailbox, 0x0088) #define USBS_RMWA USBS_ADDRREG(RxMailbox, 0x008C) // There are additional counter registers from offset 0x100 onwards. // These registers are not used by the driver, and anyway may not be // available on all hardware. // The General Mode register USBS_GMR #define USBS_GMR_VT (0x01 << 23) #define USBS_GMR_FA_MASK (0x7F << 16) #define USBS_GMR_FA_SHIFT 16 #define USBS_GMR_SOFINTVL_MASK (0x00FF << 8) #define USBS_GMR_SOFINTVL_SHIFT 8 #define USBS_GMR_SOFINTVL_DEFAULT_VALUE (0x18 << 8) #define USBS_GMR_AU (0x01 << 2) #define USBS_GMR_LE (0x01 << 1) #define USBS_GMR_RR (0x01 << 0) // The Frame Number/Version register #define USBS_VER_UVER_MASK (0x0FFFF << 16) #define USBS_VER_UVER_SHIFT 16 #define USBS_VER_UFNR_MASK (0x03FF << 0) #define USBS_VER_UFNR_SHIFT 0 // General status register 1 #define USBS_GSR1_GSR2 (0x01 << 31) #define USBS_GSR1_TMF (0x01 << 23) #define USBS_GSR1_RMF (0x01 << 22) #define USBS_GSR1_RPE2 (0x01 << 21) #define USBS_GSR1_RPE1 (0x01 << 20) #define USBS_GSR1_RPE0 (0x01 << 19) #define USBS_GSR1_RPA2 (0x01 << 18) #define USBS_GSR1_RPA1 (0x01 << 17) #define USBS_GSR1_RPA0 (0x01 << 16) #define USBS_GSR1_DER (0x01 << 10) #define USBS_GSR1_EP2FO (0x01 << 9) #define USBS_GSR1_EP1FU (0x01 << 8) #define USBS_GSR1_EP6RF (0x01 << 7) #define USBS_GSR1_EP5TF (0x01 << 6) #define USBS_GSR1_EP4RF (0x01 << 5) #define USBS_GSR1_EP3TF (0x01 << 4) #define USBS_GSR1_EP2RF (0x01 << 3) #define USBS_GSR1_EP1TF (0x01 << 2) #define USBS_GSR1_EP0RF (0x01 << 1) #define USBS_GSR1_EP0TF (0x01 << 0) // The Interrupt mask 1 bits correspond to the GSR1 bits above // General status register 2 #define USBS_GSR2_FW (0x01 << 21) #define USBS_GSR2_IFN (0x01 << 20) #define USBS_GSR2_IEA (0x01 << 19) #define USBS_GSR2_URSM (0x01 << 18) #define USBS_GSR2_URST (0x01 << 17) #define USBS_GSR2_USPD (0x01 << 16) #define USBS_GSR2_EP2OS (0x01 << 7) #define USBS_GSR2_EP2ED (0x01 << 6) #define USBS_GSR2_EP2ND (0x01 << 5) #define USBS_GSR2_EP1NT (0x01 << 4) #define USBS_GSR2_EP1ET (0x01 << 3) #define USBS_GSR2_EP1ND (0x01 << 2) #define USBS_GSR2_ES (0x01 << 1) #define USBS_GSR2_SL (0x01 << 0) // Interrupt mask 2 bits correspond to GSR2 // Endpoint control registers. // EP0 - control messages #define EP0_CR_EP0EN (0x01 << 31) #define EP0_CR_ISS (0x01 << 20) #define EP0_CR_INAK (0x01 << 19) #define EP0_CR_OSS (0x01 << 18) #define EP0_CR_NHSK0 (0x01 << 17) #define EP0_CR_ONAK (0x01 << 16) #define EP0_CR_MAXP0_MASK (0x7F << 0) #define EP0_CR_MAXP0_SHIFT 0 // EP1 - isochronous transmits #define EP1_CR_EP1EN (0x01 << 31) #define EP1_CR_TM1 (0x01 << 19) #define EP1_CR_TM1_MASK (0x01 << 19) #define EP1_CR_TM1_SZLP (0x00 << 19) #define EP1_CR_TM1_NZLP (0x01 << 19) #define EP1_CR_MAXP1_MASK (0x3FF << 0) #define EP1_CR_MAXP1_SHIFT 0 // EP2 - isochronous receives #define EP2_CR_EP2EN (0x01 << 31) #define EP2_CR_RM2_MASK (0x03 << 19) #define EP2_CR_RM2_NORMAL (0x00 << 19) #define EP2_CR_RM2_ASSEMBLE (0x02 << 19) #define EP2_CR_RM2_SEPARATE (0x03 << 19) #define EP2_CR_MAXP2_MASK (0x3FF << 0) #define EP2_CR_MAXP2_SHIFT 0 // EP3 - bulk transmits #define EP3_CR_EP3EN (0x01 << 31) #define EP3_CR_TM3 (0x01 << 19) #define EP3_CR_TM3_MASK (0x01 << 19) #define EP3_CR_TM3_SZLP (0x00 << 19) #define EP3_CR_TM3_NZLP (0x01 << 19) #define EP3_CR_SS3 (0x01 << 18) #define EP3_CR_NAK3 (0x01 << 16) #define EP3_CR_MAXP3_MASK (0x7F << 0) #define EP3_CR_MAXP3_SHIFT 0 // EP4 - bulk receives #define EP4_CR_EP4EN (0x01 << 31) #define EP4_CR_RM4_MASK (0x03 << 19) #define EP4_CR_RM4_NORMAL (0x00 << 19) #define EP4_CR_RM4_ASSEMBLE (0x02 << 19) #define EP4_CR_RM4_SEPARATE (0x03 << 19) #define EP4_CR_SS4 (0x01 << 18) #define EP4_CR_NHSK4 (0x01 << 17) #define EP4_CR_NAK4 (0x01 << 16) #define EP4_CR_MAXP4_MASK (0x7F << 0) #define EP4_CR_MAXP4_SHIFT 0 // EP5 - interrupt transmits #define EP5_CR_EP5EN (0x01 << 31) #define EP5_CR_FM (0x01 << 19) #define EP5_CR_SS5 (0x01 << 18) #define EP5_CR_NAK5 (0x01 << 16) #define EP5_CR_MAXP5_MASK (0x7F << 0) #define EP5_CR_MAXP5_SHIFT 0 // EP6 - interrupt receives #define EP6_CR_EP6EN (0x01 << 31) #define EP6_CR_SS6 (0x01 << 18) #define EP6_CR_NHSK6 (0x01 << 17) #define EP6_CR_NAK6 (0x01 << 16) #define EP6_CR_MAXP6_MASK (0x7F << 0) #define EP6_CR_MAXP6_SHIFT 0 // Some bits which can be applied to multiple transmit or receive // endpoint control registers, thus avoiding unnecessary code // duplication. These will not work for the isochronous endpoints // because those are just too special. #define EPtx_CR_EpxEN (0x01 << 31) #define EPtx_CR_SSx (0x01 << 18) #define EPtx_CR_NAKx (0x01 << 16) #define EPtx_CR_MAXPx_MASK (0x7F << 0) #define EPtx_CR_MAXPx_SHIFT 0 #define EPrx_CR_EPxEN (0x01 << 31) #define EPrx_CR_SSx (0x01 << 18) #define EPrx_CR_NHSKx (0x01 << 17) #define EPrx_CR_NAKx (0x01 << 16) #define EPrx_CR_MAXPx_MASK (0x7F << 0) #define EPrx_CR_MAXPx_SHIFT 0 // USB command register #define USBS_CMR_BUSY (0x01 << 31) #define USBS_CMR_COMMAND_MASK (0x07 << 24) #define USBS_CMR_COMMAND_SHIFT 24 #define USBS_CMR_COMMAND_TX_EP0 (0x00 << 24) #define USBS_CMR_COMMAND_TX_EP1 (0x01 << 24) #define USBS_CMR_COMMAND_TX_EP3 (0x02 << 24) #define USBS_CMR_COMMAND_TX_EP5 (0x03 << 24) #define USBS_CMR_COMMAND_ADD_POOL0 (0x04 << 24) #define USBS_CMR_COMMAND_ADD_POOL1 (0x05 << 24) #define USBS_CMR_COMMAND_ADD_POOL2 (0x06 << 24) #define USBS_CMR_SIZE_MASK (0x0FFFF << 0) #define USBS_CMR_SIZE_SHIFT 0 // TX Endpoint status #define USBS_TEPSR_EP5TS_MASK (0x03 << 24) #define USBS_TEPSR_EP5TS_SHIFT 24 #define USBS_TEPSR_EP5TS_IDLE (0x00 << 24) #define USBS_TEPSR_EP5TS_ONE (0x01 << 24) #define USBS_TEPSR_EP5TS_TWO (0x02 << 24) #define USBS_TEPSR_EP3TS_MASK (0x03 << 16) #define USBS_TEPSR_EP3TS_SHIFT 16 #define USBS_TEPSR_EP3TS_IDLE (0x00 << 16) #define USBS_TEPSR_EP3TS_ONE (0x01 << 16) #define USBS_TEPSR_EP3TS_TWO (0x02 << 16) #define USBS_TEPSR_EP1TS_MASK (0x03 << 8) #define USBS_TEPSR_EP1TS_SHIFT 8 #define USBS_TEPSR_EP1TS_IDLE (0x00 << 8) #define USBS_TEPSR_EP1TS_ONE (0x01 << 8) #define USBS_TEPSR_EP1TS_TWO (0x02 << 8) #define USBS_TEPSR_EP0TS_MASK (0x03 << 0) #define USBS_TEPSR_EP0TS_SHIFT 0 #define USBS_TEPSR_EP0TS_IDLE (0x00 << 0) #define USBS_TEPSR_EP0TS_ONE (0x01 << 0) #define USBS_TEPSR_EP0TS_TWO (0x02 << 0) // Receive pools. The RP0IR, RP1IR and RP2IR registers // all use the same bits. #define USBS_RPxIR_AL_MASK (0x07 << 28) #define USBS_RPxIR_AL_SHIFT 28 #define USBS_RPxIR_AL_NONE (0 << 28) #define USBS_RPxIR_RNOD_MASK (0x0FFFF << 0) #define USBS_RPxIR_RNOD_SHIFT 0 // The other registers do not have special bits. // Data transfers involve buffer descriptors and mailboxes. The // relevant data structures and fields need to be defined. For now // assume 32-bit mode of operation, i.e. there will be no padding // between two successive 32-bit entities // A transmit packet directory consists of up to 255 buffer // descriptors. Each buffer descriptor specifies a buffer and a size // of up to 64K. typedef struct TxBufferDescriptor { cyg_uint32 control; void* buffer; } TxBufferDescriptor; #define TXBUFDESC_CTRL_LAST (0x01 << 31) #define TXBUFDESC_CTRL_BUFDESC_MASK (0x01 << 30) #define TXBUFDESC_CTRL_BUFDESC_SHIFT 30 #define TXBUFDESC_CTRL_BUFDESC_LINK (0x00 << 30) #define TXBUFDESC_CTRL_BUFDESC_BUFDESC (0x01 << 30) #define TXBUFDESC_CTRL_SIZE_MASK (0x0FFFF << 0) #define TXBUFDESC_CTRL_SIZE_SHIFT 0 // The result of a transmit operation gets written to a mailbox // structure in memory. typedef struct TxMailbox { cyg_uint32 status; } TxMailbox; #define TXMBOX_STATUS_IBUS_ERROR (0x01 << 10) #define TXMBOX_STATUS_UNDERRUN (0x01 << 9) #define TXMBOX_STATUS_MODE_MASK (0x01 << 8) #define TXMBOX_STATUS_MODE_SHIFT 8 #define TXMBOX_STATUS_MODE_SZLP (0x00 << 8) #define TXMBOX_STATUS_MODE_NZLP (0x01 << 8) #define TXMBOX_STATUS_EPN_MASK (0x07 << 0) #define TXMBOX_STATUS_EPN_SHIFT 0 #define TXMBOX_STATUS_EPN_EP0 (0x00 << 0) #define TXMBOX_STATUS_EPN_EP1 (0x02 << 0) #define TXMBOX_STATUS_EPN_EP3 (0x04 << 0) #define TXMBOX_STATUS_EPN_EP5 (0x06 << 0) // Now for receive operations. This involves adding buffer descriptors // to one of three pools. The pools are managed by registers. typedef struct RxBufferDescriptor { cyg_uint32 control; void* buffer; } RxBufferDescriptor; #define RXBUFDESC_CTRL_LAST (0x01 << 31) #define RXBUFDESC_CTRL_BUFDESC_MASK (0x01 << 30) #define RXBUFDESC_CTRL_BUFDESC_SHIFT 30 #define RXBUFDESC_CTRL_BUFDESC_LINK (0x00 << 30) #define RXBUFDESC_CTRL_BUFDESC_BUFDESC (0x01 << 30) #define RXBUFDESC_CTRL_SIZE_MASK (0x0FFFF << 0) #define RXBUFDESC_CTRL_SIZE_SHIFT 0 typedef struct RxMailbox { cyg_uint32 status; void* address; } RxMailbox; #define RXMBOX_STATUS_EPN_MASK (0x07 << 29) #define RXMBOX_STATUS_EPN_SHIFT 29 #define RXMBOX_STATUS_EPN_EP0 (0x01 << 29) #define RXMBOX_STATUS_EPN_EP2 (0x03 << 29) #define RXMBOX_STATUS_EPN_EP4 (0x05 << 29) #define RXMBOX_STATUS_EPN_EP6 (0x07 << 29) #define RXMBOX_STATUS_CORRUPTION (0x01 << 25) #define RXMBOX_STATUS_IBUS_ERROR (0x01 << 24) #define RXMBOX_STATUS_SETUP_MASK (0x01 << 23) #define RXMBOX_STATUS_SETUP_SHIFT 23 #define RXMBOX_STATUS_SETUP_NORMAL (0x00 << 23) #define RXMBOX_STATUS_SETUP_SETUP (0x01 << 23) #define RXMBOX_STATUS_OVERRUN (0x01 << 22) #define RXMBOX_STATUS_DATA_TOGGLE (0x01 << 21) #define RXMBOX_STATUS_CRC (0x01 << 20) #define RXMBOX_STATUS_BIT_STUFFING (0x01 << 19) #define RXMBOX_STATUS_64K (0x01 << 18) #define RXMBOX_STATUS_MODE_MASK (0x03 << 16) #define RXMBOX_STATUS_MODE_SHIFT 16 #define RXMBOX_STATUS_MODE_NORMAL (0x00 << 16) #define RXMBOX_STATUS_MODE_NORMAL2 (0x01 << 16) #define RXMBOX_STATUS_MODE_ASSEMBLE (0x02 << 16) #define RXMBOX_STATUS_MODE_SEPARATE (0x03 << 16) #define RXMBOX_STATUS_SIZE_MASK (0x0FFFF << 0) #define RXMBOX_STATUS_SIZE_SHIFT 0 // ---------------------------------------------------------------------------- // Hardware work around - see NEC erratum S1, CPU to IBUS write restriction. // Reading back from the USB device after every write prevents any problems. // Strictly speaking it is only necessary to do this after every three // writes, but if there is concurrent ethernet activity then doing it // after eveyr write is safer. The frame number/version register seems // like a good one to read back from. #ifdef CYGIMP_DEVS_USB_UPD985XX_IBUS_WRITE_LIMIT # define FLUSH_IBUS() \ CYG_MACRO_START \ (void)*USBS_VER; \ CYG_MACRO_END #else # define FLUSH_IBUS() CYG_EMPTY_STATEMENT #endif // ---------------------------------------------------------------------------- // Static data. There is a data structure for each endpoint. The // implementation is essentially a private class that inherits from // common classes for control and data endpoints, but device drivers // are supposed to be written in C so some ugliness is required. // // Devtab entries are defined in usbs_upd985xx_data.cxx to make sure // that the linker does not garbage-collect them. // Support for the interrupt handling code. static cyg_interrupt usbs_upd985xx_intr_data; static cyg_handle_t usbs_upd985xx_intr_handle; // The various bits in the two interrupt status registers are read-once, // i.e. reading the register clears the bits. Since much of the processing // is deferred to DSR level, it is necessary to keep track of pending // interrupts in separate variables. If another interrupt happens during // DSR processing, these variables will be updated. The main DSR loops // until there are no interesting bits left. Interrupts have to be // disabled briefly when clearing bits. static volatile cyg_uint32 usbs_upd985xx_gsr1 = 0; static volatile cyg_uint32 usbs_upd985xx_gsr2 = 0; // Many of the interrupt bits are of no interest and it is convenient // to mask them out in the ISR, thus avoiding unnecessary dsr // invocations. static cyg_uint32 usbs_upd985xx_gsr1_mask = 0; static cyg_uint32 usbs_upd985xx_gsr2_mask = 0; // Sizes for the receive and transmit mboxes. // NOTE: it is not clear what the optimal size for these // mailboxes is. For receives maybe one per rx endpoint, // plus a spare. For transmits maybe just two, since only // one transmit at a time is supported. Mailboxes are // relatively small, so for now four each should be ok. #define RXMBOX_COUNT 4 #define TXMBOX_COUNT 4 // There is one instance of this data structure. It is allocated // in kseg0 cached memory, but during initialization a separate // pointer value is set to the kseg1 uncached equivalent. This // makes it easier to point the hardware at uncached memory without // having to worry about cache line boundaries everywhere. typedef struct uncached_data { // This partial cacheline does not actually store any data. // However it ensures that the data does not share a cacheline // with some other static, with updates to that other static // causing funny side effects on the uncached data. There is a // memory optimisation of subtracting sizeof(RxMailbox.status), // i.e. exploit knowledge of alignment. unsigned char cacheline_start[HAL_DCACHE_LINE_SIZE - sizeof(cyg_uint32)]; RxMailbox rx_mboxes[RXMBOX_COUNT]; TxMailbox tx_mboxes[TXMBOX_COUNT]; // For transmits a single buffer descriptor per endpoint suffices. // If transmit locking is enabled then actually a single buffer // descriptor for the whole system would suffice. TxBufferDescriptor ep0_tx_bufdesc; #ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 TxBufferDescriptor ep3_tx_bufdesc; #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 TxBufferDescriptor ep5_tx_bufdesc; #endif // More buffer descriptors are needed than might be expected, see // the start_rx routines. RxBufferDescriptor ep0_rx_bufdescs[4]; #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 RxBufferDescriptor ep4_rx_bufdescs[8]; #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 // Space for the start and end of a transfer, avoiding problems // with invalidating partial cache lines. unsigned char ep4_head[HAL_DCACHE_LINE_SIZE]; unsigned char ep4_tail[HAL_DCACHE_LINE_SIZE]; #endif // The "big" buffers come last, reducing the offsets for the previous // structures. It is not clear this really matters for MIPS. // // Endpoint 0 receive and transmit buffers. A transmit buffer is // convenient because the hardware pretty much expects all of the // data to be in contiguous memory, as opposed to the normal eCos // USB driver model with refill buffers etc. An alternative // implementation would keep the data in separate areas but would // require lots of TxBufferDescriptors, so in memory terms the // overheads of a single transmit buffer are not as big as might // seem. It might be possible to get things working eight bytes // at a time since the hardware appears to depend on zero-byte // terminating packets in places, but that has not been attempted. // // A separate receive buffer is useful because it can be placed in // uncached memory, avoiding the need for invalidation and // worrying about other data in the cache lines. Note that this // buffer may also get used for endpoint 6 interrupt receives // because the two endpoints share a single pool. unsigned char ep0_rx_buffer[CYGNUM_DEVS_USB_UPD985XX_EP0_RXBUFSIZE]; unsigned char ep0_tx_buffer[CYGNUM_DEVS_USB_UPD985XX_EP0_TXBUFSIZE]; // Another cacheline to prevent overlap with other statics. // This has to be full-sized since the previous field is only byte-aligned. unsigned char cacheline_end[HAL_DCACHE_LINE_SIZE]; } uncached_data; // This data structure is quite large so making it all uninitialized // means a potentially big saving in ROM-booting systems. This // requires additional effort by the endpoint initialization routines. static uncached_data cached_copy; static uncached_data* uncached = (uncached_data*)0; // Endpoint 0. See the description below. static void usbs_upd985xx_ep0_start(usbs_control_endpoint*); static void usbs_upd985xx_poll(usbs_control_endpoint*); typedef struct ep0_impl { usbs_control_endpoint common; cyg_bool rx_expecting_data; cyg_bool rx_indicator_valid; RxMailbox rx_indicator; cyg_bool tx_indicator_valid; TxMailbox tx_indicator; cyg_bool tx_needs_zero_transfer; cyg_uint32 tx_size; } ep0_impl; static ep0_impl ep0 = { common: { state: USBS_STATE_POWERED, // The hardware does not distinguish between detached, attached and powered. enumeration_data: (usbs_enumeration_data*) 0, start_fn: &usbs_upd985xx_ep0_start, poll_fn: &usbs_upd985xx_poll, interrupt_vector: CYGNUM_HAL_INTERRUPT_USB, control_buffer: { 0, 0, 0, 0, 0, 0, 0, 0 }, state_change_fn: (void (*)(usbs_control_endpoint*, void*, usbs_state_change, int)) 0, state_change_data: (void*) 0, standard_control_fn: (usbs_control_return (*)(usbs_control_endpoint*, void*)) 0, standard_control_data: (void*) 0, class_control_fn: (usbs_control_return (*)(usbs_control_endpoint*, void*)) 0, class_control_data: (void*) 0, vendor_control_fn: (usbs_control_return (*)(usbs_control_endpoint*, void*)) 0, vendor_control_data: (void*) 0, reserved_control_fn: (usbs_control_return (*)(usbs_control_endpoint*, void*)) 0, reserved_control_data: (void*) 0, buffer: (unsigned char*) 0, buffer_size: 0, fill_buffer_fn: (void (*)(usbs_control_endpoint*)) 0, fill_data: (void*) 0, fill_index: 0, complete_fn: (usbs_control_return (*)(usbs_control_endpoint*, int)) 0 }, rx_expecting_data: false, rx_indicator_valid: false, rx_indicator: { 0, (void*) 0 }, tx_indicator_valid: false, tx_indicator: { 0 }, tx_needs_zero_transfer: 0 }; extern usbs_control_endpoint usbs_upd985xx_ep0 __attribute__((alias ("ep0"))); // Endpoint 1, isochronous transmits. This endpoint is not yet // supported. Although the interface for bulk transmits should be // mostly re-usable, there are some additional error conditions if // either the host or the target fails to achieve the desired // throughput. // Endpoint 2, isochronous receives. Not yet supported for now, just // like endpoint 1. // Endpoints 3 and 5 can share some code. #if defined(CYGPKG_DEVS_USB_UPD985XX_EP3) || defined(CYGPKG_DEVS_USB_UPD985XX_EP5) // Endpoint 3, bulk transmits, and endpoint 5, either interrupt transmits // or emulation of bulk transmits. The hardware does most // of the work. typedef struct ep35_impl { usbs_tx_endpoint common; cyg_bool tx_indicator_valid; TxMailbox tx_indicator; int send_command; volatile cyg_uint32* cr; TxBufferDescriptor* tx_bufdesc; } ep35_impl; # ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 static void ep3_start_tx(usbs_tx_endpoint*); static void ep3_set_halted(usbs_tx_endpoint*, cyg_bool); static ep35_impl ep3 = { common: { start_tx_fn: &ep3_start_tx, set_halted_fn: &ep3_set_halted, complete_fn: (void (*)(void*, int)) 0, complete_data: (void*) 0, buffer: (const unsigned char*) 0, buffer_size: 0, halted: 0, }, tx_indicator_valid: false, tx_indicator: { 0 }, send_command: USBS_CMR_COMMAND_TX_EP3, cr: EP3_CR, tx_bufdesc: 0 // Needs run-time initialization }; extern usbs_tx_endpoint usbs_upd985xx_ep3 __attribute__ ((alias ("ep3"))); # endif # ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 static void ep5_start_tx(usbs_tx_endpoint*); static void ep5_set_halted(usbs_tx_endpoint*, cyg_bool); static ep35_impl ep5 = { common: { start_tx_fn: &ep5_start_tx, set_halted_fn: &ep5_set_halted, complete_fn: (void (*)(void*, int)) 0, complete_data: (void*) 0, buffer: (const unsigned char*) 0, buffer_size: 0, halted: 0, }, tx_indicator_valid: false, tx_indicator: { 0 }, send_command: USBS_CMR_COMMAND_TX_EP5, cr: EP5_CR, tx_bufdesc: 0 // Needs run-time initialization }; extern usbs_tx_endpoint usbs_upd985xx_ep5 __attribute__ ((alias ("ep5"))); # endif #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 // Endpoint 4, bulk receives. Again the hardware does the hard work. // Receive pool 2 is reserved for this endpoint. typedef struct ep4_impl { usbs_rx_endpoint common; cyg_uint32 head_size; cyg_uint32 direct_size; cyg_uint32 tail_size; cyg_bool rx_indicator_valid; RxMailbox rx_indicator; cyg_int32 tail_index; } ep4_impl; static void ep4_start_rx(usbs_rx_endpoint*); static void ep4_set_halted(usbs_rx_endpoint*, cyg_bool); static ep4_impl ep4 = { common: { start_rx_fn: &ep4_start_rx, set_halted_fn: &ep4_set_halted, complete_fn: (void (*)(void*, int)) 0, complete_data: (void*) 0, buffer: (unsigned char*) 0, buffer_size: 0, halted: 0, }, rx_indicator_valid: false, rx_indicator: { 0, (void*) 0 }, tail_index: -1 }; extern usbs_rx_endpoint usbs_upd985xx_ep4 __attribute__((alias ("ep4"))); #endif // Endpoint 6, interrupt receives. Not yet implemented. There may // be conflicts because the hardware is shared with endpoint 0. // ---------------------------------------------------------------------------- // Mailbox support. // // The transmit and receive mailboxes are shared between the // appropriate endpoints. This causes some complications if e.g. // transmits on several endpoints complete at the same time. For // example the tx mailbox might contain send indicators for endpoints // 3 and 0, but the DSR code will process endpoint 0 before endpoint // 3. // // This device driver works on the basis that there can be only one // transmit and/or receive in progress for any given endpoint, so the // relevant information can be extracted from the mailbox and put into // the per-endpoint structures. The routines below can be used to // move data from the mailboxes. They will be called in DSR context // so there is no need to worry about locking. static void drain_tx_mailbox(void) { TxMailbox* tmra = IBUS_SWAPPTR(TxMailbox, *USBS_TMRA); TxMailbox* tmwa = IBUS_SWAPPTR(TxMailbox, *USBS_TMWA); if (tmra != tmwa) { do { TxMailbox mbox = *tmra; tmra++; if (tmra == &(uncached->tx_mboxes[TXMBOX_COUNT])) { tmra = &(uncached->tx_mboxes[0]); } switch(mbox.status & TXMBOX_STATUS_EPN_MASK) { case TXMBOX_STATUS_EPN_EP0: CYG_ASSERT(false == ep0.tx_indicator_valid, "Only one ep0 transmit should be in progress at a time"); ep0.tx_indicator = mbox; ep0.tx_indicator_valid = true; break; #ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 case TXMBOX_STATUS_EPN_EP3: CYG_ASSERT(false == ep3.tx_indicator_valid, "Only one ep3 transmit should be in progress at a time"); ep3.tx_indicator = mbox; ep3.tx_indicator_valid = true; break; #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 case TXMBOX_STATUS_EPN_EP5: CYG_ASSERT(false == ep5.tx_indicator_valid, "Only one ep5 transmit should be in progress at a time"); ep5.tx_indicator = mbox; ep5.tx_indicator_valid = true; break; #endif default: break; } } while (tmra != tmwa); *USBS_TMRA = IBUS_SWAPPTR(TxMailbox, tmra); FLUSH_IBUS(); } } static void drain_rx_mailbox(void) { RxMailbox* rmra = IBUS_SWAPPTR(RxMailbox, *USBS_RMRA); RxMailbox* rmwa = IBUS_SWAPPTR(RxMailbox, *USBS_RMWA); if (rmra != rmwa) { do { RxMailbox mbox = *rmra; rmra++; if (rmra == &(uncached->rx_mboxes[RXMBOX_COUNT])) { rmra = &(uncached->rx_mboxes[0]); } switch(mbox.status & RXMBOX_STATUS_EPN_MASK) { case RXMBOX_STATUS_EPN_EP0: // Ignore zero-byte transfers. It is not clear why // these happen, but they have been observed. if (0 != (mbox.status & RXMBOX_STATUS_SIZE_MASK)) { CYG_ASSERT(false == ep0.rx_indicator_valid, "Only one ep0 receive should be in progress at a time"); ep0.rx_indicator = mbox; ep0.rx_indicator_valid = true; } break; #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 case RXMBOX_STATUS_EPN_EP4: // If an error occurs then the hardware may report // multiple rx completions, each with an IBUS error // indicator. For now only the last rx indicator is // taken into account, which means we could lose // a successful receive that happens to be followed // by an error. // NOTE: any possibility of improving on this? #if 1 CYG_ASSERT(false == ep4.rx_indicator_valid, "Only one ep4 receive should be in progress at a time"); #endif ep4.rx_indicator = mbox; ep4.rx_indicator_valid = true; break; #endif default: break; } } while (rmra != rmwa); *USBS_RMRA = IBUS_SWAPPTR(RxMailbox, rmra); FLUSH_IBUS(); } } // ---------------------------------------------------------------------------- // Transmit locking. // // According to NEC errata U3 and U4 the hardware may exhibit // undesirable behaviour if there are concurrent transmissions. There // are various ways of resolving this, but the simplest is to perform // locking in software so that at most one transmit endpoint is in use // at any one time. This approach works fine if transmissions only // involve one tx endpoint plus the control endpoint because the // control endpoint generally only gets used during initialization and // the other endpoint only gets used after initialization. If multiple // transmit endpoints are used then locking in software becomes less // acceptable, especially if isochronous transfers are used because // timing is important for those. // // There is a theoretical problem if e.g. there is a very large bulk // transfer on a busy bus and it is necessary to respond to a control // message. The control reply would be delayed, possibly causing a // violation of the USB standard and a timeout on the host. #ifdef CYGIMP_DEVS_USB_UPD985XX_SERIALIZE_TRANSMITS static void ep0_start_tx(void); # if defined(CYGPKG_DEVS_USB_UPD985XX_EP3) || defined(CYGPKG_DEVS_USB_UPD985XX_EP5) static void ep35_start_tx(ep35_impl*); # endif static cyg_bool tx_in_progress = false; static cyg_bool ep0_tx_pending = false; # ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 static cyg_bool ep3_tx_pending = false; # endif # ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 static cyg_bool ep5_tx_pending = false; # endif // Invoked from ep?_start_tx(). Scheduling may or may not be locked. static cyg_bool tx_try_lock(cyg_bool* which) { cyg_bool result; cyg_drv_dsr_lock(); if (tx_in_progress) { result = false; *which = true; } else { result = true; tx_in_progress = true; } cyg_drv_dsr_unlock(); return result; } // Invoked only from dsr context. static void tx_unlock(void) { tx_in_progress = false; if (ep0_tx_pending) { ep0_tx_pending = false; ep0_start_tx(); return; } # ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 if (ep3_tx_pending) { ep3_tx_pending = false; ep35_start_tx(&ep3); return; } # endif # ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 if (ep5_tx_pending) { ep5_tx_pending = false; ep35_start_tx(&ep5); return; } # endif } # define TX_TRY_LOCK(_x_) tx_try_lock(_x_) # define TX_UNLOCK() tx_unlock() #else # define TX_TRY_LOCK(_x_) 1 # define TX_UNLOCK() CYG_EMPTY_STATEMENT #endif // ---------------------------------------------------------------------------- // Endpoint 0 // // As usual, control messages are more complicated than the rest. // // 1) during initialization a receive is initiated into the common // eight-byte buffer, used for the standard header of the control // packet. Until that header has been received and analysed, // there is no way of knowing whether or not the host will be // sending any more data. // // 2) the control packet may indicate that the host will be sending // more data. A higher-level handler for the control message should // have provided a suitable buffer, so a receive can be started // into that buffer. A flag indicates whether we are currently // receiving a new control packet or additional data. // // 3) the host may decide to cancel that extra data and send a new // control message instead. There is a flag to indicate that // the transfer included a SETUP token. // // 4) transmits only happen when the control packet involves returning // data. Unfortunately there is a problem in that, with eCos, the // return data will generally not be in a single contiguous buffer. // Discontinuous data could be handled by having a separate buffer // descriptor for each bit of data, but it is not known in advance // how many buffer descriptors might be needed so allocating // those statically presents a problem as well. Instead a single // static buffer is used, and data from higher-level code is copied // there. This introduces a new problem: how big should that buffer // be? A configuration option is used for that. // // If endpoint 6 is in use as well then things get more complicated // because a single receive pool will be shared between endpoints 0 // and 6, and when adding a buffer to a pool there is no way of // specifying the endpoint. Hence it will be necessary to receive // into a static buffer and then copy into either an endpoint 0 or // and endpoint 6 buffer. // Fill the transmit buffer by repeatedly invoking the refill function // and copying into the ep0 tx buffer. The relevant fields in the // ep0 structure are cleared immediately and the completion function // is called, even though the data has not actually gone out. That avoids // a possible race condition where the host sends a new control packet // immediately, before the transmit-complete has been processed // (unlikely in practice, not least because ep0_tx_dsr() will get called // before ep0_rx_dsr()). static int ep0_fill_txbuffer(void) { int filled = 0; while (filled < CYGNUM_DEVS_USB_UPD985XX_EP0_TXBUFSIZE) { if (0 != ep0.common.buffer_size) { if ((filled + ep0.common.buffer_size) < CYGNUM_DEVS_USB_UPD985XX_EP0_TXBUFSIZE) { memcpy(&(uncached->ep0_tx_buffer[filled]), ep0.common.buffer, ep0.common.buffer_size); filled += ep0.common.buffer_size; ep0.common.buffer_size = 0; } else { break; } } else if ((void (*)(usbs_control_endpoint*))0 != ep0.common.fill_buffer_fn) { (*ep0.common.fill_buffer_fn)(&ep0.common); } else { break; } } CYG_ASSERT((0 == ep0.common.buffer_size) && ((void (*)(usbs_control_endpoint*))0 == ep0.common.fill_buffer_fn), \ "Endpoint 0 transmit buffer overflow"); if ((usbs_control_return (*)(usbs_control_endpoint*, int))0 != ep0.common.complete_fn) { (*ep0.common.complete_fn)(&ep0.common, 0); } ep0.common.buffer = (unsigned char*) 0; ep0.common.buffer_size = 0; ep0.common.fill_buffer_fn = 0; ep0.common.complete_fn = 0; return filled; } // Start a new receive operation on endpoint 0. This needs to happen // from a number of places, including from initialization. // // IMHO the hardware is somewhat overengineered here. All that is // needed is to receive a single eight-byte control packet, or a // small amount of additional control data. That could be achieved // by using a single buffer descriptor in the uncached structure, // plus a suitably-sized static uncached ep0_rx_buffer. // // But no, buffer descriptors must be linked and new buffers must // be added to the end. When a control packet arrives, the // receive pool continues to point at the old buffer descriptor. // So we need two buffer descriptors plus two links, switching // between them as appropriate. // // It is not at all clear what would happen if another packet // started to happen while things were being updated. There is // also potential confusion between endpoint 0 and endpoint 6 // receives. static void ep0_start_rx(cyg_uint32 size) { // The buffer descriptor to be added. This will be either // ep0_rxbufdescs[0] or ep0_rxbufdescs[2]; RxBufferDescriptor* desc = &(uncached->ep0_rx_bufdescs[0]); CYG_ASSERTC(size > 0); // Block interrupts for the duration. This does not prevent // problems if the hardware sees another packet and starts // doing things, but should prevent some software race // conditions. cyg_drv_isr_lock(); // We are about to start a new rx operation, so the // current indicator may get invalidated. ep0.rx_indicator_valid = false; // Start by looking at the current pool0 status. There are // three possibilities: during init or after reset, the pool // will be empty; otherwise the pool should point at either // rx_bufdescs[0] or rx_bufdescs[2], corresponding to the // last received packet. if (0 == (*USBS_RP0IR & USBS_RPxIR_RNOD_MASK)) { // Nothing currently in the pool. Use ep0_rx_bufdescs[0], // and no need to update a link. } else if (desc == *USBS_RP0AR) { // The pool already points at bufdescs[0], switch to bufdescs[2], // and link from bufdescs[1]. desc = &(uncached->ep0_rx_bufdescs[2]); uncached->ep0_rx_bufdescs[1].buffer = (void*) desc; } else { // The pool should point at bufdescs[2], stick with bufdescs[0] CYG_ASSERT(&(uncached->ep0_rx_bufdescs[2]) == *USBS_RP0AR, "Endpoint 0 rx buffer confusion"); uncached->ep0_rx_bufdescs[3].buffer = (void*) desc; } // Now fill in the buffer directory being added desc[0].control = RXBUFDESC_CTRL_LAST | RXBUFDESC_CTRL_BUFDESC_BUFDESC | size; desc[0].buffer = (void*) uncached->ep0_rx_buffer; desc[1].control = RXBUFDESC_CTRL_BUFDESC_LINK; desc[1].buffer = 0; while (0 != (*USBS_CMR & IBUS_SWAP32(USBS_CMR_BUSY))) { // Do nothing: this situation should be short-lived. } *USBS_CA = IBUS_SWAPPTR(void, desc); FLUSH_IBUS(); *USBS_CMR = IBUS_SWAP32(USBS_CMR_COMMAND_ADD_POOL0 | 1); FLUSH_IBUS(); cyg_drv_isr_unlock(); } // Ditto for transmits. The data is assumed to be in // uncached->ep0_tx_buffer already. A size of 0 indicates // a need to send a terminating packet explicitly. static void ep0_start_tx(void) { if (!TX_TRY_LOCK(&ep0_tx_pending)) { return; } uncached->ep0_tx_bufdesc.buffer = uncached->ep0_tx_buffer; uncached->ep0_tx_bufdesc.control = TXBUFDESC_CTRL_LAST | TXBUFDESC_CTRL_BUFDESC_BUFDESC | ep0.tx_size; cyg_drv_isr_lock(); while (0 != (*USBS_CMR & IBUS_SWAP32(USBS_CMR_BUSY))) { // Do nothing: this situation should be short-lived. } *USBS_CA = IBUS_SWAPPTR(void, &(uncached->ep0_tx_bufdesc)); FLUSH_IBUS(); *USBS_CMR = IBUS_SWAP32(USBS_CMR_COMMAND_TX_EP0 | ep0.tx_size); FLUSH_IBUS(); cyg_drv_isr_unlock(); } // An endpoint 0 transmission has completed. Usually the only action // that is needed is to drain the tx mailbox entry, otherwise it is // possible that we could end up with ep0 transmits using up all // available slots. The endpoint 0 hardware requires no further // attention, and as far as higher-level code is concerned the // transmission completed a long time ago when ep0_fill_txbuffer() // called the completion function. // // There is one special case. If the host asked for e.g. a string // descriptor and asked for 255 bytes, but the string was only // e.g. 32 bytes, then there is a problem. With a default value // for CYGNUM_DEVS_USB_UPD985XX_EP0_PKTSIZE, the data will be // transferred as four 8-byte packets, but it is necessary to // terminate the transfer with a 0-byte packet. Endpoint 0 always // operates in NZLP mode so the hardware will never generate // this last packet. Instead it is necessary to set up an // additional transfer of zero bytes. That could be done at the // same time as the main data transfer, but then it would be // necessary to poll the hardware and wait until it has finished // processing that initial transfer. static void ep0_tx_dsr(void) { if (!ep0.tx_indicator_valid) { drain_tx_mailbox(); if (!ep0.tx_indicator_valid) { // A transmit interrupt when there does not appear to be // any data? CYG_FAIL("EP0 tx DSR invoked when there is no valid tx indicator"); return; } } // There is not actually anything worth looking at in the status. ep0.tx_indicator_valid = false; if (ep0.tx_needs_zero_transfer) { ep0.tx_needs_zero_transfer = false; uncached->ep0_tx_bufdesc.buffer = uncached->ep0_tx_buffer; uncached->ep0_tx_bufdesc.control = TXBUFDESC_CTRL_LAST | TXBUFDESC_CTRL_BUFDESC_BUFDESC | 0; cyg_drv_isr_lock(); while (0 != (*USBS_CMR & IBUS_SWAP32(USBS_CMR_BUSY))) { // Do nothing: this situation should be short-lived. } *USBS_CA = IBUS_SWAPPTR(void, &(uncached->ep0_tx_bufdesc)); FLUSH_IBUS(); *USBS_CMR = IBUS_SWAP32(USBS_CMR_COMMAND_TX_EP0 | 0); FLUSH_IBUS(); cyg_drv_isr_unlock(); } else { TX_UNLOCK(); } } // An endpoint 0 receive has completed. This could be a new control // message. Or it could be the data for a previous control message. Or // it could be a new control message when expecting the data from // a previous one. The ep0.rx_expecting_data field indicates // whether or not a new control message is expected. // // At times an interrupt triggers and there is an rx indication for a // zero-byte transfer. Such a transfer may be followed immediately by // a real transfer. It is not understood why the zero-byte transfer // occurs. They are ignored by the drain_rx_mailbox() code, to make // sure that there is at most one valid rx indicator at a time. static void ep0_rx_dsr(void) { // Start by checking the rx indicator to make sure that a packet // really has been received. if (!ep0.rx_indicator_valid) { drain_rx_mailbox(); if (!ep0.rx_indicator_valid) { // Do not assert, in case of a spurious interrupt for a // zero-byte transfer. return; } } // We have a valid receive, with the data held in uncached->ep0_rx_buffer. // Are we expecting the remaining data of a control transfer? if (ep0.rx_expecting_data) { // Was this data interrupted by a new setup packet? if (0 != (ep0.rx_indicator.status & RXMBOX_STATUS_SETUP_SETUP)) { // NOTE: it is not clear from the documentation exactly what // happens here, e.g. is it guaranteed that the new control // packet appears at the start of the buffer rather than // after any data previously received? Given typical // USB host-side implementations this scenario is considered // sufficiently unlikely that no further investigation has // been carried out. // Inform higher-level code that the receive has been aborted. if ((usbs_control_return (*)(usbs_control_endpoint*, int)) 0 != ep0.common.complete_fn) { (*ep0.common.complete_fn)(&ep0.common, -EIO); } ep0.rx_expecting_data = false; ep0.common.buffer = (unsigned char*) 0; ep0.common.buffer_size = 0; ep0.common.fill_buffer_fn = 0; ep0.common.complete_fn = (usbs_control_return (*)(usbs_control_endpoint*, int)) 0; // Fall through the main control message handling code below. } else { // Data was expected and received. Transfer the data to the // user's buffer, and perform completion. usbs_control_return result; cyg_uint32 size = ep0.rx_indicator.status & RXMBOX_STATUS_SIZE_MASK; CYG_ASSERT( (usbs_control_return (*)(usbs_control_endpoint*, int))0 != ep0.common.complete_fn, \ "A completion function should be provided for OUT control messages"); CYG_ASSERT(size == ep0.common.buffer_size, "Inconsistency between buffer and transfer sizes"); memcpy(ep0.common.buffer, uncached->ep0_rx_buffer, size); result = (*ep0.common.complete_fn)(&ep0.common, 0); ep0.common.buffer = (unsigned char*) 0; ep0.common.buffer_size = 0; ep0.common.complete_fn = (usbs_control_return (*)(usbs_control_endpoint*, int)) 0; ep0.rx_expecting_data = false; // Start another receive for the next control message. // Note that there has been a window where there was no receive // in progress for endpoint 0, even though according to the // USB spec a device must always be able to accept new // control messages. ep0_start_rx(8); return; } } // When we get here we should have an eight-byte control message // in uncached->ep0_rx_buffer. This should get moved into // the ep0.common.control_buffer so that higher-level code sees // it in the appropriate location. CYG_ASSERT((ep0.rx_indicator.address == &(uncached->ep0_rx_bufdescs[0])) || \ (ep0.rx_indicator.address == &(uncached->ep0_rx_bufdescs[2])), \ "Received ep0 data should involve the ep0 rx buffer descriptor"); CYG_ASSERT(8 == (ep0.rx_indicator.status & RXMBOX_STATUS_SIZE_MASK), "Control messages should be 8 bytes"); memcpy(ep0.common.control_buffer, uncached->ep0_rx_buffer, 8); // If we have received a control packet then any reset signals really // will have come from the host and must be processed normally. // Make sure that reset interrupts are no longer masked off. if (0 == (*USBS_IMR2 & IBUS_SWAP32(USBS_GSR2_URST))) { *USBS_IMR2 |= IBUS_SWAP32(USBS_GSR2_URST); FLUSH_IBUS(); usbs_upd985xx_gsr2_mask |= USBS_GSR2_URST; } { usbs_control_return result = USBS_CONTROL_RETURN_UNKNOWN; usb_devreq* req = (usb_devreq*) ep0.common.control_buffer; int length, direction, protocol, recipient; // Now we need to do some decoding of the data. A non-zero // length field indicates that there will be a subsequent // IN or OUT phase. The direction is controlled by the // top bit of the first byte. The protocol is determined // by other bits of the top byte. length = (req->length_hi << 8) | req->length_lo; direction = req->type & USB_DEVREQ_DIRECTION_MASK; protocol = req->type & USB_DEVREQ_TYPE_MASK; recipient = req->type & USB_DEVREQ_RECIPIENT_MASK; DBG(("ep0, new control request: type %x, code %x\n", req->type, req->request)); DBG((" %s, length %d, value hi %x lo %x, index hi %x lo %x\n", (USB_DEVREQ_DIRECTION_OUT == direction) ? "out" : "in", length, req->value_hi, req->value_lo, req->index_hi, req->index_lo)); if (USB_DEVREQ_TYPE_STANDARD == protocol) { // First see if the request can be handled entirely in // this module. if (USB_DEVREQ_SET_ADDRESS == req->request) { // The USB device address should be in value_lo. // No more data is expected. int old_state = ep0.common.state; int address = req->value_lo; if ((0 != length) || (address > 127)) { result = USBS_CONTROL_RETURN_STALL; } else { *USBS_GMR = (*USBS_GMR & ~(USBS_GMR_FA_MASK | USBS_GMR_VT)) | (address << USBS_GMR_FA_SHIFT); FLUSH_IBUS(); result = USBS_CONTROL_RETURN_HANDLED; } // Switch to addressed state, informing higher-level // code of this. if (USBS_STATE_ADDRESSED != (old_state & USBS_STATE_MASK)) { ep0.common.state = USBS_STATE_ADDRESSED; if ((void (*)(usbs_control_endpoint*, void*, usbs_state_change, int))0 != ep0.common.state_change_fn) { (*ep0.common.state_change_fn)(&ep0.common, ep0.common.state_change_data, USBS_STATE_CHANGE_ADDRESSED, old_state); } } // End of SET_ADDRESS handling } else if (USB_DEVREQ_GET_STATUS == req->request) { // GET_STATUS on the device as a whole is used to // check the remote-wakeup and self-powered bits. // GET_STATUS on an endpoint is used to determine // the halted condition. // GET_STATUS on anything else has to be left to // other code. if (USB_DEVREQ_RECIPIENT_DEVICE == recipient) { // The host should expect two bytes back. if ((2 == length) && (USB_DEVREQ_DIRECTION_IN == direction)) { ep0.common.control_buffer[0] = 0; // Not self-powered, no remote wakeup ep0.common.control_buffer[1] = 0; ep0.common.buffer = ep0.common.control_buffer; ep0.common.buffer_size = 2; result = USBS_CONTROL_RETURN_HANDLED; } else { result = USBS_CONTROL_RETURN_STALL; } } else if (USB_DEVREQ_RECIPIENT_ENDPOINT == recipient) { if ((2 == length) && (USB_DEVREQ_DIRECTION_IN == direction)) { int endpoint = req->index_lo; if (0 == endpoint) { // get-status on endpoint 0 is either undefined or always valid. // endpoint 0 is always up. ep0.common.control_buffer[0] = 0; result = USBS_CONTROL_RETURN_HANDLED; } #ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 else if (((USB_DEVREQ_INDEX_DIRECTION_IN | 3) == endpoint) && (USBS_STATE_CONFIGURED == (ep0.common.state & USBS_STATE_MASK))) { ep0.common.control_buffer[0] = ep3.common.halted; result = USBS_CONTROL_RETURN_HANDLED; } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 else if (((USB_DEVREQ_INDEX_DIRECTION_OUT | 4) == endpoint) && (USBS_STATE_CONFIGURED == (ep0.common.state & USBS_STATE_MASK))) { ep0.common.control_buffer[0] = ep4.common.halted; result = USBS_CONTROL_RETURN_HANDLED; } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 else if (((USB_DEVREQ_INDEX_DIRECTION_IN | 5) == endpoint) && (USBS_STATE_CONFIGURED == (ep0.common.state & USBS_STATE_MASK))) { ep0.common.control_buffer[0] = ep5.common.halted; result = USBS_CONTROL_RETURN_HANDLED; } #endif else { // An invalid endpoint has been specified or the // endpoint can only be examined in configured state. result = USBS_CONTROL_RETURN_STALL; } if (USBS_CONTROL_RETURN_HANDLED == result) { ep0.common.control_buffer[1] = 0; ep0.common.buffer = ep0.common.control_buffer; ep0.common.buffer_size = 2; } } else { result = USBS_CONTROL_RETURN_STALL; } } // Endpoint or device get-status } else if (USB_DEVREQ_CLEAR_FEATURE == req->request) { // CLEAR_FEATURE operates in much the same way as // GET_STATUS if (USB_DEVREQ_RECIPIENT_DEVICE == recipient) { // No data should be transferred, and only remote-wakeup can be cleared. if ((0 != length) || (USB_DEVREQ_FEATURE_DEVICE_REMOTE_WAKEUP != req->value_lo)) { result = USBS_CONTROL_RETURN_STALL; } else { // Clearing remote-wakeup is a no-op. result = USBS_CONTROL_RETURN_HANDLED; } } else if (USB_DEVREQ_RECIPIENT_ENDPOINT == recipient) { // The only feature that can be cleared is endpoint-halt, no data should be transferred. if ((0 != length) || (USB_DEVREQ_FEATURE_ENDPOINT_HALT != req->value_lo)) { result = USBS_CONTROL_RETURN_STALL; } else { int endpoint = req->index_lo; if (0 == endpoint) { // Clearing halt on endpoint 0 is always a no-op since that endpoint cannot be halted } #ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 else if (((USB_DEVREQ_INDEX_DIRECTION_IN | 3) == endpoint) && (USBS_STATE_CONFIGURED == (ep0.common.state & USBS_STATE_MASK))) { ep3_set_halted(&ep3.common, false); result = USBS_CONTROL_RETURN_HANDLED; } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 else if (((USB_DEVREQ_INDEX_DIRECTION_OUT | 4) == endpoint) && (USBS_STATE_CONFIGURED == (ep0.common.state & USBS_STATE_MASK))) { ep4_set_halted(&ep4.common, false); result = USBS_CONTROL_RETURN_HANDLED; } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 else if (((USB_DEVREQ_INDEX_DIRECTION_IN | 5) == endpoint) && (USBS_STATE_CONFIGURED == (ep0.common.state & USBS_STATE_MASK))) { ep5_set_halted(&ep5.common, false); result = USBS_CONTROL_RETURN_HANDLED; } #endif else { // Invalid endpoint or not in configured state. result = USBS_CONTROL_RETURN_STALL; } } } // Endpoing or device clear-feature } else if (USB_DEVREQ_SET_FEATURE == req->request) { // SET_FEATURE also operates in much the same way as // GET_STATUS if (USB_DEVREQ_RECIPIENT_DEVICE == recipient) { // The only valid feature that can be set is remote-wakeup, // which is not supported by this driver. result = USBS_CONTROL_RETURN_STALL; } else if (USB_DEVREQ_RECIPIENT_ENDPOINT == recipient) { // Only the halt condition can be set, and no data should be transferred. // Halting endpoint 0 should probably be disallowed although the // standard does not explicitly say so. if ((0 != length) || (USB_DEVREQ_FEATURE_ENDPOINT_HALT != req->value_lo) || (USBS_STATE_CONFIGURED != (ep0.common.state & USBS_STATE_MASK))) { result = USBS_CONTROL_RETURN_STALL; } else { int endpoint = req->index_lo; if (0) { } #ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 else if ((USB_DEVREQ_INDEX_DIRECTION_IN | 3) == endpoint) { ep3_set_halted(&ep3.common, true); result = USBS_CONTROL_RETURN_HANDLED; } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 else if ((USB_DEVREQ_INDEX_DIRECTION_OUT | 4) == endpoint) { ep4_set_halted(&ep4.common, true); result = USBS_CONTROL_RETURN_HANDLED; } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 else if ((USB_DEVREQ_INDEX_DIRECTION_IN | 5) == endpoint) { ep5_set_halted(&ep5.common, true); result = USBS_CONTROL_RETURN_HANDLED; } #endif else { result = USBS_CONTROL_RETURN_STALL; } } } // Endpoint or device set-feature } // If the result has not been handled yet, pass it to // the installed callback function (if any). if (USBS_CONTROL_RETURN_UNKNOWN == result) { if ((usbs_control_return (*)(usbs_control_endpoint*, void*))0 != ep0.common.standard_control_fn) { result = (*ep0.common.standard_control_fn)(&ep0.common, ep0.common.standard_control_data); } } // If the result has still not been handled, leave it to // the default implementation in the USB slave common place. if (USBS_CONTROL_RETURN_UNKNOWN == result) { result = usbs_handle_standard_control(&ep0.common); } } else { // The other three types of control message can be // handled by similar code. usbs_control_return (*callback_fn)(usbs_control_endpoint*, void*); void* callback_arg; if (USB_DEVREQ_TYPE_CLASS == protocol) { callback_fn = ep0.common.class_control_fn; callback_arg = ep0.common.class_control_data; } else if (USB_DEVREQ_TYPE_VENDOR == protocol) { callback_fn = ep0.common.vendor_control_fn; callback_arg = ep0.common.vendor_control_data; } else { callback_fn = ep0.common.reserved_control_fn; callback_arg = ep0.common.reserved_control_data; } if ((usbs_control_return (*)(usbs_control_endpoint*, void*)) 0 == callback_fn) { result = USBS_CONTROL_RETURN_STALL; } else { result = (*callback_fn)(&ep0.common, callback_arg); } } if (USBS_CONTROL_RETURN_HANDLED != result) { // This control request cannot be handled. Generate a stall. // These stalls will be cleared automaticaly by the next // setup packet. *EP0_CR |= (EP0_CR_ISS | EP0_CR_OSS); FLUSH_IBUS(); // Start a receive for the next control message ep0_start_rx(8); } else { // The control request has been handled. Is there any more // data to be transferred? if (0 == length) { // Definitely start a receive for another control message ep0_start_rx(8); // This operation is complete so we need to ack. It // appears that the way to achieve this is to send a // zero-byte packet. ep0.tx_size = 0; ep0_start_tx(); } else { // Time to check the direction. if (USB_DEVREQ_DIRECTION_OUT == direction) { // The host expects to send more data. Higher-level code // should have provided an appropriate buffer. CYG_ASSERT( (unsigned char*) 0 != ep0.common.buffer, "A receive buffer should have been provided"); CYG_ASSERT( (usbs_control_return (*)(usbs_control_endpoint*, int))0 != ep0.common.complete_fn, \ "A completion function should be provided for OUT control messages"); CYG_ASSERT(length <= CYGNUM_DEVS_USB_UPD985XX_EP0_RXBUFSIZE, "Insufficient buffer space configured"); ep0.rx_expecting_data = true; ep0_start_rx(length); } else { // The host expects to be able to read some data. // This needs to go into a single contiguous // buffer, and then the transfer can be started. // Care has to be taken with various boundary conditions. int actual_length = ep0_fill_txbuffer(); if (actual_length > length) { actual_length = length; } if ((length != actual_length) && (0 == (actual_length % CYGNUM_DEVS_USB_UPD985XX_EP0_PKTSIZE))) { ep0.tx_needs_zero_transfer = true; } else { ep0.tx_needs_zero_transfer = false; } ep0.tx_size = actual_length; ep0_start_tx(); // And make sure that there is another receive in progress // for the next setup packet. ep0_start_rx(8); } } } // Control message handled } } // Endpoint 0 initialization also takes care of initializing generic bits // of the USB controller, for example letting through resume and suspend // interrupts and setting up the mailboxes. Also, it is necessary to // start a receive operation so that the first control message can // be processed. This code gets called during device driver initialization // and after a reset from the host. static void ep0_init(void) { // Reset the various fields in the ep0 structure. ep0.common.buffer = (unsigned char*) 0; ep0.common.buffer_size = 0; ep0.common.fill_buffer_fn = (void (*)(usbs_control_endpoint*)) 0; ep0.common.fill_data = (void*) 0; ep0.common.fill_index = 0; ep0.common.complete_fn = (usbs_control_return (*)(usbs_control_endpoint*, int)) 0; ep0.rx_expecting_data = false; ep0.tx_indicator_valid = false; ep0.rx_indicator_valid = false; ep0.tx_needs_zero_transfer = false; #ifdef CYGIMP_DEVS_USB_UPD985XX_SERIALIZE_TRANSMITS tx_in_progress = false; ep0_tx_pending = false; # ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 ep3_tx_pending = false; # endif # ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 ep5_tx_pending = false; # endif #endif // The general mode register. We do not have an address yet. The // SOFINTVL field needs to be set to its default value. The other // bits should be zero for now. *USBS_GMR = USBS_GMR_SOFINTVL_DEFAULT_VALUE; FLUSH_IBUS(); // The version register and the status registers are read-only. // Interrupt masks. Endpoint 0 transmits and receives both have // to be detected, as do the control operations. There should // be no need to worry about full mailboxes or empty receive // pools. DMA errors might be of interest, but it is not clear // what to do about them since there does not appear to be // a way of figuring out which transfer is affected. Frame number // and addressing problems are ignored, there is nothing obvious // that can be done about these. The other endpoints have their // own initialization routines. // // Care has to be taken with reset interrupts. With some hardware // the usb lines may be left floating during initialization, so // the chip believes it sees continuous reset interrupts. There // also appear to be problems if the host does generate a real // reset signal, with interrupt storms lasting 10 or more // milliseconds and preventing any other activity from taking // place. What is done here is that reset interrupts are enabled // if in the initial POWERED state. When a reset is detected, // either a spurious one or a real reset from the host, // handle_reset() will move the target to DEFAULT state, call // ep0_init() again, and reset interrupts will be masked out. // When a real control request is received from the host we // know we have a good connection and the reset interrupt will // be unmasked in ep0_rx_dsr(), so further resets from the // host will be processed correctly. If the target is disconnected // then we may again get a spurious reset interrupt, so we end // up back in DEFAULT state and the reset interrupt would be // masked again. *USBS_IMR1 = IBUS_SWAP32(USBS_GSR1_GSR2 | USBS_GSR1_EP0TF | USBS_GSR1_EP0RF); FLUSH_IBUS(); usbs_upd985xx_gsr1_mask = (USBS_GSR1_EP0TF | USBS_GSR1_EP0RF); if (USBS_STATE_DEFAULT == (ep0.common.state & USBS_STATE_MASK)) { *USBS_IMR2 = IBUS_SWAP32(USBS_GSR2_URSM | USBS_GSR2_USPD); FLUSH_IBUS(); usbs_upd985xx_gsr2_mask = (USBS_GSR2_URSM | USBS_GSR2_USPD); } else { *USBS_IMR2 = IBUS_SWAP32(USBS_GSR2_URSM | USBS_GSR2_URST | USBS_GSR2_USPD); FLUSH_IBUS(); usbs_upd985xx_gsr2_mask = (USBS_GSR2_URSM | USBS_GSR2_URST | USBS_GSR2_USPD); } // Writing to the command register is a bad idea, because even // writing 0 constitutes a command. Similarly there is no point // in writing to the command address register. // The endpoint status register is read-only. // Set the rx pool information registers to disable alerts. *USBS_RP0IR = IBUS_SWAP32(USBS_RPxIR_AL_NONE); FLUSH_IBUS(); *USBS_RP1IR = IBUS_SWAP32(USBS_RPxIR_AL_NONE); FLUSH_IBUS(); *USBS_RP2IR = IBUS_SWAP32(USBS_RPxIR_AL_NONE); FLUSH_IBUS(); // The pool address registers are read-only. The documentation // that describes initialization says that these registers need to // be filled in, but that seems wrong: providing receive buffers // involves the command register. Presumably on early revisions it // was necessary to fill in the address register. // Sort out the mailboxes. *USBS_TMSA = IBUS_SWAPPTR(TxMailbox, &(uncached->tx_mboxes[0])); FLUSH_IBUS(); *USBS_TMBA = IBUS_SWAPPTR(TxMailbox, &(uncached->tx_mboxes[TXMBOX_COUNT])); FLUSH_IBUS(); *USBS_RMSA = IBUS_SWAPPTR(RxMailbox, &(uncached->rx_mboxes[0])); FLUSH_IBUS(); *USBS_RMBA = IBUS_SWAPPTR(RxMailbox, &(uncached->rx_mboxes[RXMBOX_COUNT])); FLUSH_IBUS(); // It is not clear whether these registers actually need to be initialized. // The documentation suggests that they do, unlike TMWA and RMWA which // are taken care of by the hardware. #if 0 *USBS_TMRA = IBUS_SWAPPTR(TxMailbox, &(uncached->tx_mboxes[0])); FLUSH_IBUS(); *USBS_RMRA = IBUS_SWAPPTR(RxMailbox, &(uncached->rx_mboxes[0])); FLUSH_IBUS(); #endif // Start a receive operation for a control message. ep0_start_rx(8); // The endpoint 0 control register. The control packet size is // configurable, with a default value of 8. Setting the // enabled bit here affects the state as seen by the host. *EP0_CR = IBUS_SWAP32(EP0_CR_EP0EN | CYGNUM_DEVS_USB_UPD985XX_EP0_PKTSIZE); FLUSH_IBUS(); // The other endpoint registers will be initialized by the appropriate // _init() functions. Note that those other _init() functions should // probably be called before the ep0-enabled bit is set. } // ---------------------------------------------------------------------------- // Endpoint 1 - isochronous transmits. #if 0 // A real implementation #else static inline void ep1_init(void) { *EP1_CR = IBUS_SWAP32(0); // Clear EP1EN bit, thus disabling the endpoint FLUSH_IBUS(); } #endif // ---------------------------------------------------------------------------- // Endpoint 2 - isochronous receives #if 0 // A real implementation #else static inline void ep2_init(void) { *EP2_CR = IBUS_SWAP32(0); // Clear EP2EN bit, thus disabling the endpoint FLUSH_IBUS(); } #endif // ---------------------------------------------------------------------------- // Generic transmit support. This is intended for use with both endpoints // 3 and 5. For now the endpoint 0 code is too different. #if defined(CYGPKG_DEVS_USB_UPD985XX_EP3) || defined(CYGPKG_DEVS_USB_UPD985XX_EP5) // A utility routine for completing a transfer. This takes care of the // callback as well as resetting the buffer. static void ep35_tx_complete(ep35_impl* ep, int result) { void (*complete_fn)(void*, int) = ep->common.complete_fn; void* complete_data = ep->common.complete_data; ep->common.buffer = (unsigned char*) 0; ep->common.buffer_size = 0; ep->common.complete_fn = (void (*)(void*, int)) 0; ep->common.complete_data = (void*) 0; if ((void (*)(void*, int))0 != complete_fn) { (*complete_fn)(complete_data, result); } } static void ep35_start_tx(ep35_impl* ep) { // Is this endpoint currently stalled? If so then a size of 0 can // be used to block until the stall condition is clear, anything // else should result in an immediate callback. if (ep->common.halted) { if (0 != ep->common.buffer_size) { ep35_tx_complete(ep, -EAGAIN); } } else if (0 == ep->common.buffer_size) { // A check to see if the endpoint is halted. It isn't. ep35_tx_complete(ep, 0); } else { cyg_uint32 send_command; #if 0 diag_printf("Tx: %02x %02x %02x %02x %02x %02x %02x %02x\n", ep->common.buffer[0], ep->common.buffer[1], ep->common.buffer[2], ep->common.buffer[3], ep->common.buffer[4], ep->common.buffer[5], ep->common.buffer[6], ep->common.buffer[7]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep->common.buffer[8], ep->common.buffer[9], ep->common.buffer[10], ep->common.buffer[11], ep->common.buffer[12], ep->common.buffer[13], ep->common.buffer[14], ep->common.buffer[15]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep->common.buffer[16], ep->common.buffer[17], ep->common.buffer[18], ep->common.buffer[19], ep->common.buffer[20], ep->common.buffer[21], ep->common.buffer[22], ep->common.buffer[23]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep->common.buffer[24], ep->common.buffer[25], ep->common.buffer[26], ep->common.buffer[27], ep->common.buffer[28], ep->common.buffer[29], ep->common.buffer[30], ep->common.buffer[31]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep->common.buffer[32], ep->common.buffer[33], ep->common.buffer[34], ep->common.buffer[35], ep->common.buffer[36], ep->common.buffer[37], ep->common.buffer[38], ep->common.buffer[39]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep->common.buffer[40], ep->common.buffer[41], ep->common.buffer[42], ep->common.buffer[43], ep->common.buffer[44], ep->common.buffer[45], ep->common.buffer[46], ep->common.buffer[47]); #endif // Update the static buffer descriptor. ep->tx_bufdesc->buffer = MIPS_TO_UNCACHED(ep->common.buffer); ep->tx_bufdesc->control = ep->common.buffer_size | TXBUFDESC_CTRL_LAST | TXBUFDESC_CTRL_BUFDESC_BUFDESC; // Make sure that the entire transmit buffer is flushed to // memory. HAL_DCACHE_STORE(ep->common.buffer, ep->common.buffer_size); // Issue the send command. It is known that no transmits are // in progress for this endpoint so the upper bound of 2 // pending transmits can be ignored. The send command involves // writing to two registers in succession, so interrupts had // better be disabled while doing this. send_command = ep->send_command | ep->common.buffer_size; cyg_drv_isr_lock(); while (0 != (*USBS_CMR & IBUS_SWAP32(USBS_CMR_BUSY))) { // Do nothing: this situation should be short-lived. } *USBS_CA = IBUS_SWAPPTR(void, (void*)ep->tx_bufdesc); FLUSH_IBUS(); *USBS_CMR = IBUS_SWAP32(send_command); FLUSH_IBUS(); cyg_drv_isr_unlock(); } } // The stalled state is controlled by a single bit in the appropriate // control register. However there is a problem in that it is not // possible to abort a transmission that is already going out. // Furthermore there is no way of detecting whether or not any // packets have already gone out for this transfer: setting the halt // bit before any data has gone out is reasonably ok, doing so // in the middle of a transfer could be confusing. // // The approach taken here is to check whether or not there is a // current tx buffer. If not then the stall bit can be set immediately. // Otherwise the halted flag is set here, and it is left to the dsr // to set the stall bit when the transfer completes. This may not // be totally standards-compliant, but is probably the best solution // for now. static void ep35_set_halted(ep35_impl* ep, cyg_bool new_value) { if (ep->common.halted == new_value) { return; } // Avoid race conditions with the DSR updating the buffer fields. cyg_drv_dsr_lock(); if (new_value ){ // Set the halted flag to prevent further transmits, and if // there is no transmission currently in progress then set // the stalled bit immediately. ep->common.halted = true; if ((void*)0 == ep->common.buffer) { *(ep->cr) |= IBUS_SWAP32(EPtx_CR_SSx); FLUSH_IBUS(); } } else { // Update the hardware (that may be a no-op if the stalled bit // never got set by the DSR), and clear the halted flag. *(ep->cr) &= IBUS_SWAP32(~EPtx_CR_SSx); FLUSH_IBUS(); ep->common.halted = false; // If there is a pending request to wait until the endpoint stall // condition is clear, inform higher-level code. This test may // give false positives but those would be harmless. if (0 == ep->common.buffer_size) { ep35_tx_complete(ep, 0); } } cyg_drv_dsr_unlock(); } // An interrupt has occured related to endpoint 3 or 5 - i.e. the EP3TF // or EP5TF interrupts, nothing else seems especially relevant. It is // necessary to extract the appropriate send indicator from the tx // mailbox to determine whether or not the transmission was successful // and report status to higher-level code. static void ep35_dsr(ep35_impl* ep) { TxMailbox mbox; // Extract the transmit indicator if that has not happened // already courtesy of another DSR. if (!ep->tx_indicator_valid) { drain_tx_mailbox(); if (!ep->tx_indicator_valid) { // A transmit interrupt when there does not appear to be // any data? CYG_FAIL("ep35_dsr invoked when there is no valid tx indicator"); return; } } mbox = ep->tx_indicator; ep->tx_indicator_valid = false; #ifdef CYGIMP_DEVS_USB_UPD985XX_EP5_BULK // If emulating bulk transfers over the interrupt endpoint, and // the transfer is an exact multiple of 64 bytes, then an extra // zero-byte terminating packet needs to be sent. Care has to be // taken to do this only once. if ( (ep == &ep5) && (0 == (ep5.common.buffer_size % 64)) ) { static cyg_bool sending_zero = false; if (!sending_zero) { sending_zero = true; uncached->ep5_tx_bufdesc.buffer = uncached->ep0_tx_buffer; uncached->ep5_tx_bufdesc.control = 0 | TXBUFDESC_CTRL_LAST | TXBUFDESC_CTRL_BUFDESC_BUFDESC; cyg_drv_isr_lock(); while (0 != (*USBS_CMR & IBUS_SWAP32(USBS_CMR_BUSY))) { // Do nothing: this situation should be short-lived. } *USBS_CA = IBUS_SWAPPTR(void, (void*) &(uncached->ep5_tx_bufdesc)); FLUSH_IBUS(); *USBS_CMR = IBUS_SWAP32(USBS_CMR_COMMAND_TX_EP5 | 0); FLUSH_IBUS(); cyg_drv_isr_unlock(); // Do not complete the transfer. Instead completion has to // wait for another interrupt. return; } else { // This interrupt was for the zero-byte packet, so drop through sending_zero = false; } } #endif // If the endpoint should be halted but there was a transmit // in progress, update the hardware now. if (ep->common.halted) { *(ep->cr) |= IBUS_SWAP32(EPtx_CR_SSx); FLUSH_IBUS(); } // Allow any blocked transmits to proceed. TX_UNLOCK(); if (0 != (mbox.status & TXMBOX_STATUS_IBUS_ERROR)) { // This appears to be the only type of error that can be // detected. Possibly the transmit should be retried here // rather than reported. ep35_tx_complete(ep, -EPIPE); } else { ep35_tx_complete(ep, ep->common.buffer_size); } } #endif // Endpoints 3 or 5 // ---------------------------------------------------------------------------- // Endpoint 3 - bulk transmits. # ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 static void ep3_start_tx(usbs_tx_endpoint* endpoint) { CYG_ASSERT( endpoint == &ep3.common, "USB data transfer involves the wrong endpoint"); CYG_ASSERT( ep3.common.buffer_size < (64 * 1024), "Specified transfer size too large for current implementation"); if (TX_TRY_LOCK(&ep3_tx_pending)) { ep35_start_tx(&ep3); } } static void ep3_set_halted(usbs_tx_endpoint* endpoint, cyg_bool new_value) { CYG_ASSERT(endpoint = &ep3.common, "USB set-stall operation involves the wrong endpoint"); ep35_set_halted(&ep3, new_value); } // Initialization. This gets called during the device driver // initialization and after a reset. The main job is to initialize the // EP3 control register, but the relevant bits of the interrupt mask // are set here as well. The tx mailboxes are shared with other // endpoints, so that is handled by ep0_init(). Any traffic that // happened before the reset needs to be cleaned up. static void ep3_init(void) { // Assume 64 byte packets, terminate transfers with a zero-byte packet // if necessary since this endpoint is used for bulk transfers. *EP3_CR = IBUS_SWAP32(EP3_CR_EP3EN | EP3_CR_TM3_SZLP | 64); FLUSH_IBUS(); *USBS_IMR1 |= IBUS_SWAP32(USBS_GSR1_EP3TF); FLUSH_IBUS(); usbs_upd985xx_gsr1_mask |= IBUS_SWAP32(USBS_GSR1_EP3TF); ep3.common.halted = false; ep3.tx_indicator_valid = false; ep3.tx_bufdesc = &(uncached->ep3_tx_bufdesc); ep35_tx_complete(&ep3, -EPIPE); } #else static inline void ep3_init(void) { *EP3_CR = 0; // Clear EP3EN bit, thus disabling the endpoint FLUSH_IBUS(); } # endif // Endpoint 3 configured in // ---------------------------------------------------------------------------- // Repeat for endpoint 5 # ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 static void ep5_start_tx(usbs_tx_endpoint* endpoint) { CYG_ASSERT( endpoint == &ep5.common, "USB data transfer involves the wrong endpoint"); #ifdef CYGIMP_DEVS_USB_UPD985XX_EP5_BULK CYG_ASSERT( ep5.common.buffer_size < (64 * 1024), "Specified transfer size too large for current implementation"); #else CYG_ASSERT( ep5.common.buffer_size <= 64, "Specified transfer size too large for current implementation"); #endif if (TX_TRY_LOCK(&ep5_tx_pending)) { ep35_start_tx(&ep5); } } static void ep5_set_halted(usbs_tx_endpoint* endpoint, cyg_bool new_value) { CYG_ASSERT(endpoint = &ep5.common, "USB set-stall operation involves the wrong endpoint"); ep35_set_halted(&ep5, new_value); } // Initialization. This gets called during the device driver // initialization and after a reset. The main job is to initialize the // EP5 control register, but the relevant bits of the interrupt mask // are set here as well. The tx mailboxes are shared with other // endpoints, so that is handled by ep0_init(). Any traffic that // happened before the reset needs to be cleaned up. static void ep5_init(void) { // Assume 64 byte packets, terminate transfers with a zero-byte packet // if necessary since this endpoint is used for bulk transfers. *EP5_CR = IBUS_SWAP32(EP5_CR_EP5EN | 64); FLUSH_IBUS(); *USBS_IMR1 |= IBUS_SWAP32(USBS_GSR1_EP5TF); FLUSH_IBUS(); usbs_upd985xx_gsr1_mask |= IBUS_SWAP32(USBS_GSR1_EP5TF); ep5.common.halted = false; ep5.tx_indicator_valid = false; ep5.tx_bufdesc = &(uncached->ep5_tx_bufdesc); ep35_tx_complete(&ep5, -EPIPE); } #else static inline void ep5_init(void) { *EP5_CR = 0; // Clear EP5EN bit, thus disabling the endpoint FLUSH_IBUS(); } #endif // Endpoint 5 configured in #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 // ---------------------------------------------------------------------------- // Endpoint 4 - bulk receives. // // Bulk receives are mostly straightforward, but the cache does involve // a complication. The assumption is that the receive buffer will be in // cached memory, and is unlikely to be cacheline-aligned. Clearly the bulk // of the buffer has to be invalidated. However the receive buffer may share // some cache lines with other data at the head and tail, and invalidating // those would be wrong. // // The solution here is to split up a receive in to up to three areas, // head, main, and tail, where the head and tail are statically // allocated in uncached memory. Any one or two of these areas may be // unused, depending on alignment and transfer size. The main area // corresponds to the central section of the supplied receive buffer, // will be cacheline-aligned, and invalidated at the start of a receive. // Data will be copied from the head and tail areas into the receive // buffer by the dsr on completion of the transfer. // // There are additional complications caused by the hardware's need for // linked buffers. static void ep4_rx_complete(int result) { void (*complete_fn)(void*, int) = ep4.common.complete_fn; void* complete_data = ep4.common.complete_data; #if 0 *EP4_CR = IBUS_SWAP32(EP4_CR_EP4EN | EP4_CR_RM4_ASSEMBLE | EP4_CR_NAK4 | 64); FLUSH_IBUS(); #endif #if 0 if (result > 0) { diag_printf("Rx: %02x %02x %02x %02x %02x %02x %02x %02x\n", ep4.common.buffer[0], ep4.common.buffer[1], ep4.common.buffer[2], ep4.common.buffer[3], ep4.common.buffer[4], ep4.common.buffer[5], ep4.common.buffer[6], ep4.common.buffer[7]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep4.common.buffer[8], ep4.common.buffer[9], ep4.common.buffer[10], ep4.common.buffer[11], ep4.common.buffer[12], ep4.common.buffer[13], ep4.common.buffer[14], ep4.common.buffer[15]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep4.common.buffer[16], ep4.common.buffer[17], ep4.common.buffer[18], ep4.common.buffer[19], ep4.common.buffer[20], ep4.common.buffer[21], ep4.common.buffer[22], ep4.common.buffer[23]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep4.common.buffer[24], ep4.common.buffer[25], ep4.common.buffer[26], ep4.common.buffer[27], ep4.common.buffer[28], ep4.common.buffer[29], ep4.common.buffer[30], ep4.common.buffer[31]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep4.common.buffer[32], ep4.common.buffer[33], ep4.common.buffer[34], ep4.common.buffer[35], ep4.common.buffer[36], ep4.common.buffer[37], ep4.common.buffer[38], ep4.common.buffer[39]); diag_printf(" %02x %02x %02x %02x %02x %02x %02x %02x\n", ep4.common.buffer[40], ep4.common.buffer[41], ep4.common.buffer[42], ep4.common.buffer[43], ep4.common.buffer[44], ep4.common.buffer[45], ep4.common.buffer[46], ep4.common.buffer[47]); } #endif ep4.common.buffer = (unsigned char*) 0; ep4.common.buffer_size = 0; ep4.common.complete_fn = (void (*)(void*, int)) 0; ep4.common.complete_data = (void*) 0; if ((void (*)(void*, int))0 != complete_fn) { (*complete_fn)(complete_data, result); } } static void ep4_start_rx(usbs_rx_endpoint* ep) { CYG_ASSERT( ep == &ep4.common, "USB data transfer involves the wrong endpoint"); // Is this endpoint currently stalled? If so then a size of 0 can // be used to block until the stall condition is clear, anything // else should result in an immediate callback. if (ep4.common.halted) { if (0 != ep4.common.buffer_size) { ep4_rx_complete(-EAGAIN); } } else if (0 == ep4.common.buffer_size) { // A check to see if the endpoint is halted. It isn't. ep4_rx_complete(0); } else { // Time to work out how much data should go into the uncached // head and tail buffers, how much can go directly into the // receive buffer, how much memory needs to be invalidated, and so on. cyg_uint32 buffer_arith; // Where to start filling in buffer descriptors. cyg_uint32 first_bufdesc; // And the current buffer descriptor. cyg_uint32 current_bufdesc; CYG_ASSERT( ep4.common.buffer_size < (64 * 1024), "Specified transfer size too large for current implementation"); // If there has not been a receive operation, tail_index // will still be set to -1. Otherwise it will be somewhere // between 1 and 3, or between 5 and 7, depending on // whether the previous receive operation used the // first four buffer descriptors or the last four. if (ep4.tail_index < 4) { first_bufdesc = 4; } else { first_bufdesc = 0; } current_bufdesc = first_bufdesc; // Arithmetic, especially remainder operators, requires // integers rather than a pointer. buffer_arith = (cyg_uint32) ep4.common.buffer; // The size of the "head" area. This involves up to // (cacheline-1) bytes, so that the main receive buffer // is suitably aligned. ep4.head_size = ((buffer_arith + HAL_DCACHE_LINE_SIZE - 1) & ~(HAL_DCACHE_LINE_SIZE - 1)) - buffer_arith; if (ep4.head_size > ep4.common.buffer_size) { ep4.head_size = ep4.common.buffer_size; } if (0 < ep4.head_size) { // It is necessary to receive some data into the uncached head area. uncached->ep4_rx_bufdescs[current_bufdesc].buffer = uncached->ep4_head; uncached->ep4_rx_bufdescs[current_bufdesc].control = RXBUFDESC_CTRL_BUFDESC_BUFDESC | ep4.head_size; current_bufdesc++; } // Now for the size of the main area. This is the rest of the // transfer size, minus the tail area. ep4.direct_size = ep4.common.buffer_size - ep4.head_size; ep4.direct_size &= ~(HAL_DCACHE_LINE_SIZE - 1); if (ep4.direct_size > 0) { uncached->ep4_rx_bufdescs[current_bufdesc].buffer = MIPS_TO_UNCACHED(ep4.common.buffer + ep4.head_size); uncached->ep4_rx_bufdescs[current_bufdesc].control = RXBUFDESC_CTRL_BUFDESC_BUFDESC | ep4.direct_size; current_bufdesc++; HAL_DCACHE_INVALIDATE(ep4.common.buffer + ep4.head_size, ep4.direct_size); } // And the size of the tail. This is the transfer size minus what we have accumulated so far. ep4.tail_size = ep4.common.buffer_size - (ep4.head_size + ep4.direct_size); if (ep4.tail_size > 0) { uncached->ep4_rx_bufdescs[current_bufdesc].buffer = uncached->ep4_tail; uncached->ep4_rx_bufdescs[current_bufdesc].control = RXBUFDESC_CTRL_BUFDESC_BUFDESC | ep4.tail_size; current_bufdesc++; } // Or the LAST bit into the last of these buffer descriptors. uncached->ep4_rx_bufdescs[current_bufdesc - 1].control |= RXBUFDESC_CTRL_LAST; // Turn the current one into a link descriptor. uncached->ep4_rx_bufdescs[current_bufdesc].control = RXBUFDESC_CTRL_BUFDESC_LINK; uncached->ep4_rx_bufdescs[current_bufdesc].buffer = 0; // The buffer descriptors have now been sorted out. Time to // add this receive buffer to the pool. Atomicity becomes // important for some of these steps. cyg_drv_isr_lock(); // Update the link pointer used for the last receive operation // to point at the new set of buffer descriptors. if (-1 != ep4.tail_index) { uncached->ep4_rx_bufdescs[ep4.tail_index].buffer = (void*) &(uncached->ep4_rx_bufdescs[first_bufdesc]); } // Keep track of the link pointer used for the last receive // operation. ep4.tail_index = current_bufdesc; while (0 != (*USBS_CMR & IBUS_SWAP32(USBS_CMR_BUSY))) { // Do nothing: this situation should be short-lived. } *USBS_CA = IBUS_SWAPPTR(void, (void*)&(uncached->ep4_rx_bufdescs[first_bufdesc])); FLUSH_IBUS(); *USBS_CMR = IBUS_SWAP32(USBS_CMR_COMMAND_ADD_POOL2 | 1); FLUSH_IBUS(); #if 0 *EP4_CR = IBUS_SWAP32(EP4_CR_EP4EN | EP4_CR_RM4_ASSEMBLE | 64); FLUSH_IBUS(); #endif cyg_drv_isr_unlock(); } } // The stalled state is controlled by a single bit in the appropriate // control register. When it comes to ongoing receives, the reasoning // here is much the same as for ep3_set_halted(). Arguably this is not // quite right. set_halted() is most likely to be called in response // to a control request from the host, and the host is unlikely to // transmit any data at the same time as making this request. Hence the // host will see the wrong behaviour: it can still make one transfer // after asking for the stalled bit to be set. Since there does not // appear to be any way to cancel a supplied receive, this behaviour // still seems the most sensible. static void ep4_set_halted(usbs_rx_endpoint* ep, cyg_bool new_state) { CYG_ASSERT(ep == &ep4.common, "USB set-stall operation involves the wrong endpoint"); if (ep4.common.halted == new_state) { return; } // Avoid race conditions with the DSR updating the buffer fields. cyg_drv_dsr_lock(); if (new_state){ // Set the halted flag to prevent further transmits, and if // there is no receive currently in progress then set // the stalled bit immediately. ep4.common.halted = true; if ((void*)0 == ep4.common.buffer) { *EP4_CR |= IBUS_SWAP32(EP4_CR_SS4); FLUSH_IBUS(); } } else { // Update the hardware (that may be a no-op if the stalled bit // never got set by the DSR), and clear the halted flag. *EP4_CR &= IBUS_SWAP32(~EP4_CR_SS4); FLUSH_IBUS(); ep4.common.halted = false; // If there is a pending request to wait until the endpoint stall // condition is clear, inform higher-level code. This test may // give false positives but those would be harmless. if (0 == ep4.common.buffer_size) { ep4_rx_complete(0); } } cyg_drv_dsr_unlock(); } // An interrupt has occured related to endpoint 4 - i.e. the EP4RF // interrupt, nothing else seems especially relevant. The ISR will // have set the NAK bit in the control register. It is necessary to // extract the appropriate receive indicator from the rx mailbox to // determine whether or not the transmission was successful and how // much data was actually received, clear the interrupt bit, and // report status to higher-level code. static void ep4_dsr(void) { // Extract the transmit indicator if that has not happened // already courtesy of another DSR. if (!ep4.rx_indicator_valid) { drain_rx_mailbox(); if (!ep4.rx_indicator_valid) { // A receive interrupt when there does not appear to be // any data? It appears that this can happen when there // are error conditions. #if 1 CYG_FAIL("EP4_DSR invoked when there is no valid rx indicator"); #endif return; } } ep4.rx_indicator_valid = false; // If the endpoint should be halted but there was a transmit // in progress, update the hardware now. if (ep4.common.halted) { *EP4_CR |= IBUS_SWAP32(EP4_CR_SS4); FLUSH_IBUS(); } if (0 != (ep4.rx_indicator.status & (RXMBOX_STATUS_IBUS_ERROR | 0))) { // This appears to be the only type of error that can be // detected. Everything else gets retried by the hardware, // except when using the isochronous endpoint. ep4_rx_complete(-EPIPE); } else { cyg_uint32 actual_size = ep4.rx_indicator.status & RXMBOX_STATUS_SIZE_MASK; // Either the transfer size must match the requested size, or the // supplied buffer should have been aligned to cacheline boundaries. // Anything else risks leaving the receive pool in a confused // state and there is no way of cleaning things up. CYG_ASSERT((actual_size == ep4.common.buffer_size) || ((0 == ep4.head_size) && (0 == ep4.tail_size)), \ "Buffers should be cacheline aligned if the protocol involves partial transfers"); // If there was some data in the head, move it from uncached // to cached. Ditto for tail. Note that these copies may be // for data that has not actually been transferred if the // actual transfer is less than expected, but overwriting bits // of the receive buffer in such circumstances should be // harmless. if (ep4.head_size > 0) { memcpy(ep4.common.buffer, uncached->ep4_head, ep4.head_size); } if (ep4.tail_size > 0) { memcpy(ep4.common.buffer + ep4.head_size + ep4.direct_size, uncached->ep4_tail, ep4.tail_size); } ep4_rx_complete(actual_size); } } // Initialization. This gets called during the device driver // initialization and after a reset. The main job is to initialize the // EP4 control register, but the relevant bits of the interrupt mask // are set here as well. The rx mailboxes are shared with other // endpoints, so that is handled by ep0_init(). Any traffic that // happened before the reset needs to be cleaned up. static void ep4_init(void) { // Assume 64 byte packets, and use assemble mode so that we get a // single rx indication per transfer. In practice the buffer // directory will only ever contain one entry so there should be // no discernible difference between normal, assemble, or separate // mode. #if 0 *EP4_CR = IBUS_SWAP32(EP4_CR_EP4EN | EP4_CR_RM4_ASSEMBLE | EP4_CR_NAK4 | 64); FLUSH_IBUS(); #else *EP4_CR = IBUS_SWAP32(EP4_CR_EP4EN | EP4_CR_RM4_ASSEMBLE | 64); FLUSH_IBUS(); #endif *USBS_IMR1 |= IBUS_SWAP32(USBS_GSR1_EP4RF); FLUSH_IBUS(); usbs_upd985xx_gsr1_mask |= USBS_GSR1_EP4RF; ep4.common.halted = false; ep4.rx_indicator_valid = false; ep4.tail_index = -1; ep4_rx_complete(-EPIPE); } #else static inline void ep4_init(void) { *EP4_CR = 0; // Clear EP4EN bit, thus disabling the endpoint FLUSH_IBUS(); } #endif // Endpoint 4 configured in // ---------------------------------------------------------------------------- // Endpoint 6 - interrupt receives #if 0 // A real implementation #else static inline void ep6_init(void) { *EP6_CR = 0; // Clear EP6EN bit, thus disabling the endpoint FLUSH_IBUS(); } #endif // ---------------------------------------------------------------------------- // Make sure the hardware is in a known state by completely resetting // the USB controller. This gets called during device driver // initialization, and again whenever the host issues a reset signal. // The previous state is unknown. Even during eCos initialization // RedBoot may have involved USB I/O, or some POST code may have // performed loopback tests. The various endpoint init routines will // also perform software resets as appropriate. static void usbs_upd985xx_handle_reset(void) { // Reset the USB hardware. This involves poking the warm reset // register and then polling the matching status register. It is // assumed that this poll will take a short time, and in practice // the loop appears to terminate immediately. *S_WRCR = S_WRCR_USBWR; FLUSH_IBUS(); while (0 == (*S_WRSR & S_WRCR_USBWR)) { // Do nothing. } // Get all the endpoints into a known state - for disabled // endpoints these init calls are inlined and just disable the // relevant hardware. ep0_init(); ep1_init(); ep2_init(); ep3_init(); ep4_init(); ep5_init(); ep6_init(); } // ---------------------------------------------------------------------------- // Start(). This is typically called by the application itself once // everything else has been initialized, i.e. when the host should be // able to start talking to this device. There is no actual bit in the // chip itself to switch the USB pins from tri-state, so instead the // platform HAL has to supply appropriate functionality. In the absence // of such functionality things will only work if you start up eCos // with the USB cable disconnected, then plug in the cable once // start() has been called. // // The device driver initialization will have already set up the // hardware, and the first action from the host should be a reset // signal which will cause a re-initialization. There is no need // to do anything else here. static void usbs_upd985xx_ep0_start(usbs_control_endpoint* endpoint) { CYG_ASSERT( endpoint == &ep0.common, "USB startup involves the wrong endpoint"); // If there is additional platform-specific initialization to // perform, do it now. This macro can come from the platform HAL, // but may not be available on all platforms. #ifdef UPD985XX_USB_PLATFORM_INIT UPD985XX_USB_PLATFORM_INIT(); #endif } // ---------------------------------------------------------------------------- // The main DSR // // This gets called by the interrupt system or by the polling code // when one or more USB-related events have occurred. The ISR code // will have updated globals usbs_upd985xx_gsr1 and usbs_upd98x0x_gsr2 // to indicate which events are pending. When running in interrupt // mode it is possible that further interrupts will occur while the // DSR is running, and hence that these globals will be updated // while the DSR is running. This is handled by a loop. A side effect // is that the DSR may get called when all the work has already // been done by a previous DSR, but that is harmless. static void usbs_upd985xx_dsr(cyg_vector_t vector, cyg_ucount32 count, cyg_addrword_t data) { CYG_ASSERT(CYGNUM_HAL_INTERRUPT_USB == vector, "USB DSR should only be invoked for USB interrupts" ); CYG_ASSERT(0 == data, "The USB DSR needs no global data pointer"); while ((0 != usbs_upd985xx_gsr1) || (0 != usbs_upd985xx_gsr2)) { // Only update the globals in one place since it involves // the overhead of two function calls. cyg_uint32 gsr1, gsr2; cyg_drv_isr_lock(); gsr1 = usbs_upd985xx_gsr1; gsr2 = usbs_upd985xx_gsr2; usbs_upd985xx_gsr1 = 0; usbs_upd985xx_gsr2 = 0; cyg_drv_isr_unlock(); if (0 != gsr2) { // Treat reset specially. If there has been a reset then none of // the other bits are of any interest. if (0 != (USBS_GSR2_URST & gsr2)) { int old_state = ep0.common.state; // Update the state. ep0_init() detects this state change and // updates imr2 appropriate, preventing a continuous storm // of reset interrupts. ep0.common.state = USBS_STATE_DEFAULT; usbs_upd985xx_handle_reset(); // This state change must be reported to higher-level code if ((void (*)(usbs_control_endpoint*, void*, usbs_state_change, int))0 != ep0.common.state_change_fn) { (*ep0.common.state_change_fn)(&ep0.common, ep0.common.state_change_data, USBS_STATE_CHANGE_RESET, old_state); } break; } // There is possible confusion if both suspend and resume // bits are set. Was there a suspend, quickly followed by // a resume? Were we already suspended, then resumed, now // suspended again? For now this complication is ignored and // resume is given priority over suspend. if (0 != (USBS_GSR2_URSM & gsr2)) { int old_state = ep0.common.state; if (0 != (old_state & USBS_STATE_SUSPENDED)) { ep0.common.state &= ~USBS_STATE_SUSPENDED; if ((void (*)(usbs_control_endpoint*, void*, usbs_state_change, int))0 != ep0.common.state_change_fn) { (*ep0.common.state_change_fn)(&ep0.common, ep0.common.state_change_data, USBS_STATE_CHANGE_RESUMED, old_state); } } } else if (0 != (USBS_GSR2_USPD & gsr2)) { int old_state = ep0.common.state; if (0 == (old_state & USBS_STATE_SUSPENDED)) { ep0.common.state |= USBS_STATE_SUSPENDED; if ((void (*)(usbs_control_endpoint*, void*, usbs_state_change, int))0 != ep0.common.state_change_fn) { (*ep0.common.state_change_fn)(&ep0.common, ep0.common.state_change_data, USBS_STATE_CHANGE_SUSPENDED, old_state); } } } else { // Handle error conditions on the isochronous endpoints? } } if (0 != gsr1) { if (0 != (USBS_GSR1_EP0TF & gsr1)) { ep0_tx_dsr(); } if (0 != (USBS_GSR1_EP0RF & gsr1)) { ep0_rx_dsr(); } #if 0 // EP1FU? if (0 != (USBS_GSR1_EP1TF & gsr1)) { ep1_dsr(); } #endif #if 0 // EP2FO? if (0 != (USBS_GSR1_EP2RF & gsr1)) { ep1_dsr(); } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 if (0 != (USBS_GSR1_EP3TF & gsr1)) { ep35_dsr(&ep3); } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 if (0 != (USBS_GSR1_EP4RF & gsr1)) { ep4_dsr(); } #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 if (0 != (USBS_GSR1_EP5TF & gsr1)) { ep35_dsr(&ep5); } #endif #if 0 if (0 != (USBS_GSR1_EP6RF & gsr1)) { ep6_dsr(); } #endif } } // while there are unprocessed interrupts } // ---------------------------------------------------------------------------- // Interrupt handling. // // There are two status registers to look at. These are read-once // registers, i.e. reading the register causes all bits to be cleared, // so the relevant state has to be preserved in volatile globals which // can then be examined by the DSR. In theory the top bit of the first // status register can be used to check whether or not there is // anything of interest in the second one, but it seems quicker to // just read both registers. After masking out interrupts that are of // no interest, some global flags are updated. If this leaves a // non-zero value then the DSR must be invoked. static cyg_uint32 usbs_upd985xx_isr(cyg_vector_t vector, cyg_addrword_t data) { CYG_ASSERT(CYGNUM_HAL_INTERRUPT_USB == vector, "USB ISR should only be invoked for USB interrupts"); CYG_ASSERT(0 == data, "The UPD985xx ISR needs no global data pointer"); usbs_upd985xx_gsr1 |= IBUS_SWAP32(*USBS_GSR1) & usbs_upd985xx_gsr1_mask; usbs_upd985xx_gsr2 |= IBUS_SWAP32(*USBS_GSR2) & usbs_upd985xx_gsr2_mask; cyg_drv_interrupt_acknowledge(vector); return ((0 == usbs_upd985xx_gsr1) && (0 == usbs_upd985xx_gsr2)) ? CYG_ISR_HANDLED : CYG_ISR_CALL_DSR; } // ---------------------------------------------------------------------------- // Polling support. It is not clear that this is going to work particularly // well since according to the documentation the hardware does not generate // NAKs automatically - instead the ISR has to set the appropriate bits // sufficiently quickly to avoid confusing the host. // // Calling the isr directly avoids duplicating code, but means that // cyg_drv_interrupt_acknowledge() will get called when not inside a // real interrupt handler. This should be harmless. static void usbs_upd985xx_poll(usbs_control_endpoint* endpoint) { CYG_ASSERT(endpoint == &ep0.common, "USB poll involves the wrong endpoint"); if (CYG_ISR_CALL_DSR == usbs_upd985xx_isr(CYGNUM_HAL_INTERRUPT_USB, 0)) { usbs_upd985xx_dsr(CYGNUM_HAL_INTERRUPT_USB, 0, 0); } } // ---------------------------------------------------------------------------- // Initialization // // This routine gets called from a prioritized static constructor during // eCos startup. void usbs_upd985xx_init(void) { // Make sure the uncached data structure is accessed through // kseg1. uncached = (uncached_data*) MIPS_TO_UNCACHED(&cached_copy); // Perform a full hardware reset. usbs_upd985xx_handle_reset(); // It is possible and desirable to install the interrupt handler // here, even though there will be no interrupts for a while yet. // FIXME: is 99 a sensible interrupt priority :-? cyg_drv_interrupt_create(CYGNUM_HAL_INTERRUPT_USB, 99, // priority 0, // data &usbs_upd985xx_isr, &usbs_upd985xx_dsr, &usbs_upd985xx_intr_handle, &usbs_upd985xx_intr_data); cyg_drv_interrupt_attach(usbs_upd985xx_intr_handle); cyg_drv_interrupt_unmask(CYGNUM_HAL_INTERRUPT_USB); } // ---------------------------------------------------------------------------- // Testing support. usbs_testing_endpoint usbs_testing_endpoints[] = { { endpoint_type : USB_ENDPOINT_DESCRIPTOR_ATTR_CONTROL, endpoint_number : 0, endpoint_direction : USB_ENDPOINT_DESCRIPTOR_ENDPOINT_IN, endpoint : (void*) &ep0.common, #ifdef CYGVAR_DEVS_USB_UPD985XX_EP0_DEVTAB_ENTRY devtab_entry : CYGDAT_DEVS_USB_UPD985XX_DEVTAB_BASENAME "0c", #else devtab_entry : (const char*) 0, #endif min_size : 1, // zero-byte control transfers are meaningless #if (CYGNUM_DEVS_USB_UPD985XX_EP0_RXBUFSIZE < CYGNUM_DEVS_USB_UPD985XX_EP0_TXBUFSIZE) max_size : CYGNUM_DEVS_USB_UPD985XX_EP0_RXBUFSIZE, #else max_size : CYGNUM_DEVS_USB_UPD985XX_EP0_TXBUFSIZE, #endif max_in_padding : 0, alignment : 0 }, #ifdef CYGPKG_DEVS_USB_UPD985XX_EP3 { endpoint_type : USB_ENDPOINT_DESCRIPTOR_ATTR_BULK, endpoint_number : 3, endpoint_direction : USB_ENDPOINT_DESCRIPTOR_ENDPOINT_IN, endpoint : (void*) &ep3.common, # ifdef CYGVAR_DEVS_USB_UPD985XX_EP3_DEVTAB_ENTRY devtab_entry : CYGDAT_DEVS_USB_UPD985XX_DEVTAB_BASENAME "3w", # else devtab_entry : (const char*) 0, # endif min_size : 0, max_size : 0x0FFFF, // Driver limitation, only a single buffer descriptor is used max_in_padding : 0, alignment : HAL_DCACHE_LINE_SIZE }, #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP4 { endpoint_type : USB_ENDPOINT_DESCRIPTOR_ATTR_BULK, endpoint_number : 4, endpoint_direction : USB_ENDPOINT_DESCRIPTOR_ENDPOINT_OUT, endpoint : (void*) &ep4.common, # ifdef CYGVAR_DEVS_USB_UPD985XX_EP4_DEVTAB_ENTRY devtab_entry : CYGDAT_DEVS_USB_UPD985XX_DEVTAB_BASENAME "4r", # else devtab_entry : (const char*) 0, # endif min_size : 1, max_size : 0x0FFFF, // Driver limitation max_in_padding : 0, alignment : HAL_DCACHE_LINE_SIZE }, #endif #ifdef CYGPKG_DEVS_USB_UPD985XX_EP5 { # ifdef CYGIMP_DEVS_USB_UPD985XX_EP5_BULK endpoint_type : USB_ENDPOINT_DESCRIPTOR_ATTR_BULK, # else endpoint_type : USB_ENDPOINT_DESCRIPTOR_ATTR_INTERRUPT, # endif endpoint_number : 5, endpoint_direction : USB_ENDPOINT_DESCRIPTOR_ENDPOINT_IN, endpoint : (void*) &ep5.common, # ifdef CYGVAR_DEVS_USB_UPD985XX_EP5_DEVTAB_ENTRY devtab_entry : CYGDAT_DEVS_USB_UPD985XX_DEVTAB_BASENAME "5w", # else devtab_entry : (const char*) 0, # endif min_size : 1, max_size : 0x0FFFF, // Driver limitation, only a single buffer descriptor is used max_in_padding : 0, alignment : HAL_DCACHE_LINE_SIZE }, #endif USBS_TESTING_ENDPOINTS_TERMINATOR };
Go to most recent revision | Compare with Previous | Blame | View Log