From fc5ab020965067c457a4b5a9eb15648b6874bef2 Mon Sep 17 00:00:00 2001 From: Hariprasad Shenai Date: Fri, 27 Jun 2014 19:23:49 +0530 Subject: [PATCH] cxgb4: Replaced the backdoor mechanism to access the HW memory with PCIe Window method Rip out a bunch of redundant PCI-E Memory Window Read/Write routines, collapse the more general purpose routines into a single routine thereby eliminating the need for a large stack frame (and extra data copying) in the outer routine, change everything to use the improved routine t4_memory_rw. Based on origninal work by Casey Leedom and Steve Wise Signed-off-by: Casey Leedom Signed-off-by: Steve Wise Signed-off-by: Hariprasad Shenai Signed-off-by: David S. Miller --- drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 15 +- .../net/ethernet/chelsio/cxgb4/cxgb4_main.c | 58 +++-- drivers/net/ethernet/chelsio/cxgb4/t4_hw.c | 246 ++++++++---------- drivers/net/ethernet/chelsio/cxgb4/t4_regs.h | 1 + 4 files changed, 148 insertions(+), 172 deletions(-) diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h index 4fd215185f9..f338a7fcebf 100644 --- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h +++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h @@ -654,6 +654,7 @@ struct adapter { struct dentry *debugfs_root; spinlock_t stats_lock; + spinlock_t win0_lock ____cacheline_aligned_in_smp; }; /* Defined bit width of user definable filter tuples @@ -960,8 +961,17 @@ int t4_wait_dev_ready(struct adapter *adap); int t4_link_start(struct adapter *adap, unsigned int mbox, unsigned int port, struct link_config *lc); int t4_restart_aneg(struct adapter *adap, unsigned int mbox, unsigned int port); -int t4_memory_write(struct adapter *adap, int mtype, u32 addr, u32 len, - __be32 *buf); + +#define T4_MEMORY_WRITE 0 +#define T4_MEMORY_READ 1 +int t4_memory_rw(struct adapter *adap, int win, int mtype, u32 addr, u32 len, + __be32 *buf, int dir); +static inline int t4_memory_write(struct adapter *adap, int mtype, u32 addr, + u32 len, __be32 *buf) +{ + return t4_memory_rw(adap, 0, mtype, addr, len, buf, 0); +} + int t4_seeprom_wp(struct adapter *adapter, bool enable); int get_vpd_params(struct adapter *adapter, struct vpd_params *p); int t4_load_fw(struct adapter *adapter, const u8 *fw_data, unsigned int size); @@ -1059,7 +1069,6 @@ int t4_ofld_eq_free(struct adapter *adap, unsigned int mbox, unsigned int pf, int t4_handle_fw_rpl(struct adapter *adap, const __be64 *rpl); void t4_db_full(struct adapter *adapter); void t4_db_dropped(struct adapter *adapter); -int t4_mem_win_read_len(struct adapter *adap, u32 addr, __be32 *data, int len); int t4_fwaddrspace_write(struct adapter *adap, unsigned int mbox, u32 addr, u32 val); void t4_sge_decode_idma_state(struct adapter *adapter, int state); diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c index 524a8b2894e..74ef0e69cdc 100644 --- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c +++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c @@ -3066,6 +3066,8 @@ static ssize_t mem_read(struct file *file, char __user *buf, size_t count, loff_t avail = file_inode(file)->i_size; unsigned int mem = (uintptr_t)file->private_data & 3; struct adapter *adap = file->private_data - mem; + __be32 *data; + int ret; if (pos < 0) return -EINVAL; @@ -3074,29 +3076,24 @@ static ssize_t mem_read(struct file *file, char __user *buf, size_t count, if (count > avail - pos) count = avail - pos; - while (count) { - size_t len; - int ret, ofst; - __be32 data[16]; + data = t4_alloc_mem(count); + if (!data) + return -ENOMEM; - if ((mem == MEM_MC) || (mem == MEM_MC1)) - ret = t4_mc_read(adap, mem % MEM_MC, pos, data, NULL); - else - ret = t4_edc_read(adap, mem, pos, data, NULL); - if (ret) - return ret; + spin_lock(&adap->win0_lock); + ret = t4_memory_rw(adap, 0, mem, pos, count, data, T4_MEMORY_READ); + spin_unlock(&adap->win0_lock); + if (ret) { + t4_free_mem(data); + return ret; + } + ret = copy_to_user(buf, data, count); - ofst = pos % sizeof(data); - len = min(count, sizeof(data) - ofst); - if (copy_to_user(buf, (u8 *)data + ofst, len)) - return -EFAULT; + t4_free_mem(data); + if (ret) + return -EFAULT; - buf += len; - pos += len; - count -= len; - } - count = pos - *ppos; - *ppos = pos; + *ppos = pos + count; return count; } @@ -3757,7 +3754,11 @@ static int read_eq_indices(struct adapter *adap, u16 qid, u16 *pidx, u16 *cidx) __be64 indices; int ret; - ret = t4_mem_win_read_len(adap, addr, (__be32 *)&indices, 8); + spin_lock(&adap->win0_lock); + ret = t4_memory_rw(adap, 0, MEM_EDC0, addr, + sizeof(indices), (__be32 *)&indices, + T4_MEMORY_READ); + spin_unlock(&adap->win0_lock); if (!ret) { *cidx = (be64_to_cpu(indices) >> 25) & 0xffff; *pidx = (be64_to_cpu(indices) >> 9) & 0xffff; @@ -5076,7 +5077,7 @@ static int adap_init0_config(struct adapter *adapter, int reset) adapter->fn, 0, 1, params, val); if (ret == 0) { /* - * For t4_memory_write() below addresses and + * For t4_memory_rw() below addresses and * sizes have to be in terms of multiples of 4 * bytes. So, if the Configuration File isn't * a multiple of 4 bytes in length we'll have @@ -5092,8 +5093,9 @@ static int adap_init0_config(struct adapter *adapter, int reset) mtype = FW_PARAMS_PARAM_Y_GET(val[0]); maddr = FW_PARAMS_PARAM_Z_GET(val[0]) << 16; - ret = t4_memory_write(adapter, mtype, maddr, - size, data); + spin_lock(&adapter->win0_lock); + ret = t4_memory_rw(adapter, 0, mtype, maddr, + size, data, T4_MEMORY_WRITE); if (ret == 0 && resid != 0) { union { __be32 word; @@ -5104,10 +5106,12 @@ static int adap_init0_config(struct adapter *adapter, int reset) last.word = data[size >> 2]; for (i = resid; i < 4; i++) last.buf[i] = 0; - ret = t4_memory_write(adapter, mtype, - maddr + size, - 4, &last.word); + ret = t4_memory_rw(adapter, 0, mtype, + maddr + size, + 4, &last.word, + T4_MEMORY_WRITE); } + spin_unlock(&adapter->win0_lock); } } diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c index 2c3d00d77ef..eb5a278e804 100644 --- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c +++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c @@ -413,78 +413,41 @@ int t4_edc_read(struct adapter *adap, int idx, u32 addr, __be32 *data, u64 *ecc) return 0; } -/* - * t4_mem_win_rw - read/write memory through PCIE memory window - * @adap: the adapter - * @addr: address of first byte requested - * @data: MEMWIN0_APERTURE bytes of data containing the requested address - * @dir: direction of transfer 1 => read, 0 => write - * - * Read/write MEMWIN0_APERTURE bytes of data from MC starting at a - * MEMWIN0_APERTURE-byte-aligned address that covers the requested - * address @addr. - */ -static int t4_mem_win_rw(struct adapter *adap, u32 addr, __be32 *data, int dir) -{ - int i; - u32 win_pf = is_t4(adap->params.chip) ? 0 : V_PFNUM(adap->fn); - - /* - * Setup offset into PCIE memory window. Address must be a - * MEMWIN0_APERTURE-byte-aligned address. (Read back MA register to - * ensure that changes propagate before we attempt to use the new - * values.) - */ - t4_write_reg(adap, PCIE_MEM_ACCESS_OFFSET, - (addr & ~(MEMWIN0_APERTURE - 1)) | win_pf); - t4_read_reg(adap, PCIE_MEM_ACCESS_OFFSET); - - /* Collecting data 4 bytes at a time upto MEMWIN0_APERTURE */ - for (i = 0; i < MEMWIN0_APERTURE; i = i+0x4) { - if (dir) - *data++ = (__force __be32) t4_read_reg(adap, - (MEMWIN0_BASE + i)); - else - t4_write_reg(adap, (MEMWIN0_BASE + i), - (__force u32) *data++); - } - - return 0; -} - /** * t4_memory_rw - read/write EDC 0, EDC 1 or MC via PCIE memory window * @adap: the adapter + * @win: PCI-E Memory Window to use * @mtype: memory type: MEM_EDC0, MEM_EDC1 or MEM_MC * @addr: address within indicated memory type * @len: amount of memory to transfer * @buf: host memory buffer - * @dir: direction of transfer 1 => read, 0 => write + * @dir: direction of transfer T4_MEMORY_READ (1) or T4_MEMORY_WRITE (0) * * Reads/writes an [almost] arbitrary memory region in the firmware: the - * firmware memory address, length and host buffer must be aligned on - * 32-bit boudaries. The memory is transferred as a raw byte sequence - * from/to the firmware's memory. If this memory contains data - * structures which contain multi-byte integers, it's the callers - * responsibility to perform appropriate byte order conversions. + * firmware memory address and host buffer must be aligned on 32-bit + * boudaries; the length may be arbitrary. The memory is transferred as + * a raw byte sequence from/to the firmware's memory. If this memory + * contains data structures which contain multi-byte integers, it's the + * caller's responsibility to perform appropriate byte order conversions. */ -static int t4_memory_rw(struct adapter *adap, int mtype, u32 addr, u32 len, - __be32 *buf, int dir) +int t4_memory_rw(struct adapter *adap, int win, int mtype, u32 addr, + u32 len, __be32 *buf, int dir) { - u32 pos, start, end, offset, memoffset; - u32 edc_size, mc_size; - int ret = 0; - __be32 *data; + u32 pos, offset, resid, memoffset; + u32 edc_size, mc_size, win_pf, mem_reg, mem_aperture, mem_base; - /* - * Argument sanity checks ... + /* Argument sanity checks ... */ - if ((addr & 0x3) || (len & 0x3)) + if (addr & 0x3) return -EINVAL; - data = vmalloc(MEMWIN0_APERTURE); - if (!data) - return -ENOMEM; + /* It's convenient to be able to handle lengths which aren't a + * multiple of 32-bits because we often end up transferring files to + * the firmware. So we'll handle that by normalizing the length here + * and then handling any residual transfer at the end. + */ + resid = len & 0x3; + len -= resid; /* Offset into the region of memory which is being accessed * MEM_EDC0 = 0 @@ -505,66 +468,98 @@ static int t4_memory_rw(struct adapter *adap, int mtype, u32 addr, u32 len, /* Determine the PCIE_MEM_ACCESS_OFFSET */ addr = addr + memoffset; - /* - * The underlaying EDC/MC read routines read MEMWIN0_APERTURE bytes - * at a time so we need to round down the start and round up the end. - * We'll start copying out of the first line at (addr - start) a word - * at a time. + /* Each PCI-E Memory Window is programmed with a window size -- or + * "aperture" -- which controls the granularity of its mapping onto + * adapter memory. We need to grab that aperture in order to know + * how to use the specified window. The window is also programmed + * with the base address of the Memory Window in BAR0's address + * space. For T4 this is an absolute PCI-E Bus Address. For T5 + * the address is relative to BAR0. */ - start = addr & ~(MEMWIN0_APERTURE-1); - end = (addr + len + MEMWIN0_APERTURE-1) & ~(MEMWIN0_APERTURE-1); - offset = (addr - start)/sizeof(__be32); + mem_reg = t4_read_reg(adap, + PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_BASE_WIN, + win)); + mem_aperture = 1 << (GET_WINDOW(mem_reg) + 10); + mem_base = GET_PCIEOFST(mem_reg) << 10; + if (is_t4(adap->params.chip)) + mem_base -= adap->t4_bar0; + win_pf = is_t4(adap->params.chip) ? 0 : V_PFNUM(adap->fn); - for (pos = start; pos < end; pos += MEMWIN0_APERTURE, offset = 0) { + /* Calculate our initial PCI-E Memory Window Position and Offset into + * that Window. + */ + pos = addr & ~(mem_aperture-1); + offset = addr - pos; - /* - * If we're writing, copy the data from the caller's memory - * buffer + /* Set up initial PCI-E Memory Window to cover the start of our + * transfer. (Read it back to ensure that changes propagate before we + * attempt to use the new value.) + */ + t4_write_reg(adap, + PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_OFFSET, win), + pos | win_pf); + t4_read_reg(adap, + PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_OFFSET, win)); + + /* Transfer data to/from the adapter as long as there's an integral + * number of 32-bit transfers to complete. + */ + while (len > 0) { + if (dir == T4_MEMORY_READ) + *buf++ = (__force __be32) t4_read_reg(adap, + mem_base + offset); + else + t4_write_reg(adap, mem_base + offset, + (__force u32) *buf++); + offset += sizeof(__be32); + len -= sizeof(__be32); + + /* If we've reached the end of our current window aperture, + * move the PCI-E Memory Window on to the next. Note that + * doing this here after "len" may be 0 allows us to set up + * the PCI-E Memory Window for a possible final residual + * transfer below ... */ - if (!dir) { - /* - * If we're doing a partial write, then we need to do - * a read-modify-write ... - */ - if (offset || len < MEMWIN0_APERTURE) { - ret = t4_mem_win_rw(adap, pos, data, 1); - if (ret) - break; - } - while (offset < (MEMWIN0_APERTURE/sizeof(__be32)) && - len > 0) { - data[offset++] = *buf++; - len -= sizeof(__be32); - } + if (offset == mem_aperture) { + pos += mem_aperture; + offset = 0; + t4_write_reg(adap, + PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_OFFSET, + win), pos | win_pf); + t4_read_reg(adap, + PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_OFFSET, + win)); } - - /* - * Transfer a block of memory and bail if there's an error. - */ - ret = t4_mem_win_rw(adap, pos, data, dir); - if (ret) - break; - - /* - * If we're reading, copy the data into the caller's memory - * buffer. - */ - if (dir) - while (offset < (MEMWIN0_APERTURE/sizeof(__be32)) && - len > 0) { - *buf++ = data[offset++]; - len -= sizeof(__be32); - } } - vfree(data); - return ret; -} + /* If the original transfer had a length which wasn't a multiple of + * 32-bits, now's where we need to finish off the transfer of the + * residual amount. The PCI-E Memory Window has already been moved + * above (if necessary) to cover this final transfer. + */ + if (resid) { + union { + __be32 word; + char byte[4]; + } last; + unsigned char *bp; + int i; + + if (dir == T4_MEMORY_WRITE) { + last.word = (__force __be32) t4_read_reg(adap, + mem_base + offset); + for (bp = (unsigned char *)buf, i = resid; i < 4; i++) + bp[i] = last.byte[i]; + } else { + last.word = *buf; + for (i = resid; i < 4; i++) + last.byte[i] = 0; + t4_write_reg(adap, mem_base + offset, + (__force u32) last.word); + } + } -int t4_memory_write(struct adapter *adap, int mtype, u32 addr, u32 len, - __be32 *buf) -{ - return t4_memory_rw(adap, mtype, addr, len, buf, 0); + return 0; } #define EEPROM_STAT_ADDR 0x7bfc @@ -2528,39 +2523,6 @@ int t4_fwaddrspace_write(struct adapter *adap, unsigned int mbox, return t4_wr_mbox(adap, mbox, &c, sizeof(c), NULL); } -/** - * t4_mem_win_read_len - read memory through PCIE memory window - * @adap: the adapter - * @addr: address of first byte requested aligned on 32b. - * @data: len bytes to hold the data read - * @len: amount of data to read from window. Must be <= - * MEMWIN0_APERATURE after adjusting for 16B for T4 and - * 128B for T5 alignment requirements of the the memory window. - * - * Read len bytes of data from MC starting at @addr. - */ -int t4_mem_win_read_len(struct adapter *adap, u32 addr, __be32 *data, int len) -{ - int i, off; - u32 win_pf = is_t4(adap->params.chip) ? 0 : V_PFNUM(adap->fn); - - /* Align on a 2KB boundary. - */ - off = addr & MEMWIN0_APERTURE; - if ((addr & 3) || (len + off) > MEMWIN0_APERTURE) - return -EINVAL; - - t4_write_reg(adap, PCIE_MEM_ACCESS_OFFSET, - (addr & ~MEMWIN0_APERTURE) | win_pf); - t4_read_reg(adap, PCIE_MEM_ACCESS_OFFSET); - - for (i = 0; i < len; i += 4) - *data++ = (__force __be32) t4_read_reg(adap, - (MEMWIN0_BASE + off + i)); - - return 0; -} - /** * t4_mdio_rd - read a PHY register through MDIO * @adap: the adapter diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h index f6b2ee13e71..ae7776471ce 100644 --- a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h +++ b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h @@ -400,6 +400,7 @@ #define WINDOW_MASK 0x000000ffU #define WINDOW_SHIFT 0 #define WINDOW(x) ((x) << WINDOW_SHIFT) +#define GET_WINDOW(x) (((x) >> WINDOW_SHIFT) & WINDOW_MASK) #define PCIE_MEM_ACCESS_OFFSET 0x306c #define ENABLE (1U << 30) #define FUNCTION(x) ((x) << 12) -- 2.41.0