From 03dd4cb26d967f9588437b0fc9cc0e8353322bb7 Mon Sep 17 00:00:00 2001 From: André Fabian Silva Delgado Date: Fri, 25 Mar 2016 03:53:42 -0300 Subject: Linux-libre 4.5-gnu --- drivers/spi/spi-loopback-test.c | 1006 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 1006 insertions(+) create mode 100644 drivers/spi/spi-loopback-test.c (limited to 'drivers/spi/spi-loopback-test.c') diff --git a/drivers/spi/spi-loopback-test.c b/drivers/spi/spi-loopback-test.c new file mode 100644 index 000000000..cf4bb36be --- /dev/null +++ b/drivers/spi/spi-loopback-test.c @@ -0,0 +1,1006 @@ +/* + * linux/drivers/spi/spi-loopback-test.c + * + * (c) Martin Sperl + * + * Loopback test driver to test several typical spi_message conditions + * that a spi_master driver may encounter + * this can also get used for regression testing + * + * This program 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 of the License, or + * (at your option) any later version. + * + * This program 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. + */ + +#include +#include +#include +#include +#include +#include +#include +#include + +#include "spi-test.h" + +/* flag to only simulate transfers */ +int simulate_only; +module_param(simulate_only, int, 0); +MODULE_PARM_DESC(simulate_only, "if not 0 do not execute the spi message"); + +/* dump spi messages */ +int dump_messages; +module_param(dump_messages, int, 0); +MODULE_PARM_DESC(dump_messages, + "=1 dump the basic spi_message_structure, " \ + "=2 dump the spi_message_structure including data, " \ + "=3 dump the spi_message structure before and after execution"); +/* the device is jumpered for loopback - enabling some rx_buf tests */ +int loopback; +module_param(loopback, int, 0); +MODULE_PARM_DESC(loopback, + "if set enable loopback mode, where the rx_buf " \ + "is checked to match tx_buf after the spi_message " \ + "is executed"); + +/* run only a specific test */ +int run_only_test = -1; +module_param(run_only_test, int, 0); +MODULE_PARM_DESC(run_only_test, + "only run the test with this number (0-based !)"); + +/* the actual tests to execute */ +static struct spi_test spi_tests[] = { + { + .description = "tx/rx-transfer - start of page", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_rx_align = ITERATE_ALIGN, + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + .rx_buf = RX(0), + }, + }, + }, + { + .description = "tx/rx-transfer - crossing PAGE_SIZE", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_rx_align = ITERATE_ALIGN, + .transfers = { + { + .len = 1, + .tx_buf = TX(PAGE_SIZE - 4), + .rx_buf = RX(PAGE_SIZE - 4), + }, + }, + }, + { + .description = "tx-transfer - only", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + }, + }, + }, + { + .description = "rx-transfer - only", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_rx_align = ITERATE_ALIGN, + .transfers = { + { + .len = 1, + .rx_buf = RX(0), + }, + }, + }, + { + .description = "two tx-transfers - alter both", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(0) | BIT(1), + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + }, + { + .len = 1, + /* this is why we cant use ITERATE_MAX_LEN */ + .tx_buf = TX(SPI_TEST_MAX_SIZE_HALF), + }, + }, + }, + { + .description = "two tx-transfers - alter first", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(1), + .transfers = { + { + .len = 1, + .tx_buf = TX(64), + }, + { + .len = 1, + .tx_buf = TX(0), + }, + }, + }, + { + .description = "two tx-transfers - alter second", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(0), + .transfers = { + { + .len = 16, + .tx_buf = TX(0), + }, + { + .len = 1, + .tx_buf = TX(64), + }, + }, + }, + { + .description = "two transfers tx then rx - alter both", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(0) | BIT(1), + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + }, + { + .len = 1, + .rx_buf = RX(0), + }, + }, + }, + { + .description = "two transfers tx then rx - alter tx", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(0), + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + }, + { + .len = 1, + .rx_buf = RX(0), + }, + }, + }, + { + .description = "two transfers tx then rx - alter rx", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(1), + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + }, + { + .len = 1, + .rx_buf = RX(0), + }, + }, + }, + { + .description = "two tx+rx transfers - alter both", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(0) | BIT(1), + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + .rx_buf = RX(0), + }, + { + .len = 1, + /* making sure we align without overwrite + * the reason we can not use ITERATE_MAX_LEN + */ + .tx_buf = TX(SPI_TEST_MAX_SIZE_HALF), + .rx_buf = RX(SPI_TEST_MAX_SIZE_HALF), + }, + }, + }, + { + .description = "two tx+rx transfers - alter first", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(0), + .transfers = { + { + .len = 1, + /* making sure we align without overwrite */ + .tx_buf = TX(1024), + .rx_buf = RX(1024), + }, + { + .len = 1, + /* making sure we align without overwrite */ + .tx_buf = TX(0), + .rx_buf = RX(0), + }, + }, + }, + { + .description = "two tx+rx transfers - alter second", + .fill_option = FILL_COUNT_8, + .iterate_len = { ITERATE_MAX_LEN }, + .iterate_tx_align = ITERATE_ALIGN, + .iterate_transfer_mask = BIT(1), + .transfers = { + { + .len = 1, + .tx_buf = TX(0), + .rx_buf = RX(0), + }, + { + .len = 1, + /* making sure we align without overwrite */ + .tx_buf = TX(1024), + .rx_buf = RX(1024), + }, + }, + }, + + { /* end of tests sequence */ } +}; + +static int spi_loopback_test_probe(struct spi_device *spi) +{ + int ret; + + dev_info(&spi->dev, "Executing spi-loopback-tests\n"); + + ret = spi_test_run_tests(spi, spi_tests); + + dev_info(&spi->dev, "Finished spi-loopback-tests with return: %i\n", + ret); + + return ret; +} + +/* non const match table to permit to change via a module parameter */ +static struct of_device_id spi_loopback_test_of_match[] = { + { .compatible = "linux,spi-loopback-test", }, + { } +}; + +/* allow to override the compatible string via a module_parameter */ +module_param_string(compatible, spi_loopback_test_of_match[0].compatible, + sizeof(spi_loopback_test_of_match[0].compatible), + 0000); + +MODULE_DEVICE_TABLE(of, spi_loopback_test_of_match); + +static struct spi_driver spi_loopback_test_driver = { + .driver = { + .name = "spi-loopback-test", + .owner = THIS_MODULE, + .of_match_table = spi_loopback_test_of_match, + }, + .probe = spi_loopback_test_probe, +}; + +module_spi_driver(spi_loopback_test_driver); + +MODULE_AUTHOR("Martin Sperl "); +MODULE_DESCRIPTION("test spi_driver to check core functionality"); +MODULE_LICENSE("GPL"); + +/*-------------------------------------------------------------------------*/ + +/* spi_test implementation */ + +#define RANGE_CHECK(ptr, plen, start, slen) \ + ((ptr >= start) && (ptr + plen <= start + slen)) + +/* we allocate one page more, to allow for offsets */ +#define SPI_TEST_MAX_SIZE_PLUS (SPI_TEST_MAX_SIZE + PAGE_SIZE) + +static void spi_test_print_hex_dump(char *pre, const void *ptr, size_t len) +{ + /* limit the hex_dump */ + if (len < 1024) { + print_hex_dump(KERN_INFO, pre, + DUMP_PREFIX_OFFSET, 16, 1, + ptr, len, 0); + return; + } + /* print head */ + print_hex_dump(KERN_INFO, pre, + DUMP_PREFIX_OFFSET, 16, 1, + ptr, 512, 0); + /* print tail */ + pr_info("%s truncated - continuing at offset %04zx\n", + pre, len - 512); + print_hex_dump(KERN_INFO, pre, + DUMP_PREFIX_OFFSET, 16, 1, + ptr + (len - 512), 512, 0); +} + +static void spi_test_dump_message(struct spi_device *spi, + struct spi_message *msg, + bool dump_data) +{ + struct spi_transfer *xfer; + int i; + u8 b; + + dev_info(&spi->dev, " spi_msg@%pK\n", msg); + if (msg->status) + dev_info(&spi->dev, " status: %i\n", + msg->status); + dev_info(&spi->dev, " frame_length: %i\n", + msg->frame_length); + dev_info(&spi->dev, " actual_length: %i\n", + msg->actual_length); + + list_for_each_entry(xfer, &msg->transfers, transfer_list) { + dev_info(&spi->dev, " spi_transfer@%pK\n", xfer); + dev_info(&spi->dev, " len: %i\n", xfer->len); + dev_info(&spi->dev, " tx_buf: %pK\n", xfer->tx_buf); + if (dump_data && xfer->tx_buf) + spi_test_print_hex_dump(" TX: ", + xfer->tx_buf, + xfer->len); + + dev_info(&spi->dev, " rx_buf: %pK\n", xfer->rx_buf); + if (dump_data && xfer->rx_buf) + spi_test_print_hex_dump(" RX: ", + xfer->rx_buf, + xfer->len); + /* check for unwritten test pattern on rx_buf */ + if (xfer->rx_buf) { + for (i = 0 ; i < xfer->len ; i++) { + b = ((u8 *)xfer->rx_buf)[xfer->len - 1 - i]; + if (b != SPI_TEST_PATTERN_UNWRITTEN) + break; + } + if (i) + dev_info(&spi->dev, + " rx_buf filled with %02x starts at offset: %i\n", + SPI_TEST_PATTERN_UNWRITTEN, + xfer->len - i); + } + } +} + +struct rx_ranges { + struct list_head list; + u8 *start; + u8 *end; +}; + +int rx_ranges_cmp(void *priv, struct list_head *a, struct list_head *b) +{ + struct rx_ranges *rx_a = list_entry(a, struct rx_ranges, list); + struct rx_ranges *rx_b = list_entry(b, struct rx_ranges, list); + + if (rx_a->start > rx_b->start) + return 1; + if (rx_a->start < rx_b->start) + return -1; + return 0; +} + +static int spi_check_rx_ranges(struct spi_device *spi, + struct spi_message *msg, + void *rx) +{ + struct spi_transfer *xfer; + struct rx_ranges ranges[SPI_TEST_MAX_TRANSFERS], *r; + int i = 0; + LIST_HEAD(ranges_list); + u8 *addr; + int ret = 0; + + /* loop over all transfers to fill in the rx_ranges */ + list_for_each_entry(xfer, &msg->transfers, transfer_list) { + /* if there is no rx, then no check is needed */ + if (!xfer->rx_buf) + continue; + /* fill in the rx_range */ + if (RANGE_CHECK(xfer->rx_buf, xfer->len, + rx, SPI_TEST_MAX_SIZE_PLUS)) { + ranges[i].start = xfer->rx_buf; + ranges[i].end = xfer->rx_buf + xfer->len; + list_add(&ranges[i].list, &ranges_list); + i++; + } + } + + /* if no ranges, then we can return and avoid the checks...*/ + if (!i) + return 0; + + /* sort the list */ + list_sort(NULL, &ranges_list, rx_ranges_cmp); + + /* and iterate over all the rx addresses */ + for (addr = rx; addr < (u8 *)rx + SPI_TEST_MAX_SIZE_PLUS; addr++) { + /* if we are the DO not write pattern, + * then continue with the loop... + */ + if (*addr == SPI_TEST_PATTERN_DO_NOT_WRITE) + continue; + + /* check if we are inside a range */ + list_for_each_entry(r, &ranges_list, list) { + /* if so then set to end... */ + if ((addr >= r->start) && (addr < r->end)) + addr = r->end; + } + /* second test after a (hopefull) translation */ + if (*addr == SPI_TEST_PATTERN_DO_NOT_WRITE) + continue; + + /* if still not found then something has modified too much */ + /* we could list the "closest" transfer here... */ + dev_err(&spi->dev, + "loopback strangeness - rx changed outside of allowed range at: %pK\n", + addr); + /* do not return, only set ret, + * so that we list all addresses + */ + ret = -ERANGE; + } + + return ret; +} + +static int spi_test_check_loopback_result(struct spi_device *spi, + struct spi_message *msg, + void *tx, void *rx) +{ + struct spi_transfer *xfer; + u8 rxb, txb; + size_t i; + int ret; + + /* checks rx_buffer pattern are valid with loopback or without */ + ret = spi_check_rx_ranges(spi, msg, rx); + if (ret) + return ret; + + /* if we run without loopback, then return now */ + if (!loopback) + return 0; + + /* if applicable to transfer check that rx_buf is equal to tx_buf */ + list_for_each_entry(xfer, &msg->transfers, transfer_list) { + /* if there is no rx, then no check is needed */ + if (!xfer->rx_buf) + continue; + /* so depending on tx_buf we need to handle things */ + if (xfer->tx_buf) { + for (i = 1; i < xfer->len; i++) { + txb = ((u8 *)xfer->tx_buf)[i]; + rxb = ((u8 *)xfer->rx_buf)[i]; + if (txb != rxb) + goto mismatch_error; + } + } else { + /* first byte received */ + txb = ((u8 *)xfer->rx_buf)[0]; + /* first byte may be 0 or xff */ + if (!((txb == 0) || (txb == 0xff))) { + dev_err(&spi->dev, + "loopback strangeness - we expect 0x00 or 0xff, but not 0x%02x\n", + txb); + return -EINVAL; + } + /* check that all bytes are identical */ + for (i = 1; i < xfer->len; i++) { + rxb = ((u8 *)xfer->rx_buf)[i]; + if (rxb != txb) + goto mismatch_error; + } + } + } + + return 0; + +mismatch_error: + dev_err(&spi->dev, + "loopback strangeness - transfer missmatch on byte %04zx - expected 0x%02x, but got 0x%02x\n", + i, txb, rxb); + + return -EINVAL; +} + +static int spi_test_translate(struct spi_device *spi, + void **ptr, size_t len, + void *tx, void *rx) +{ + size_t off; + + /* return on null */ + if (!*ptr) + return 0; + + /* in the MAX_SIZE_HALF case modify the pointer */ + if (((size_t)*ptr) & SPI_TEST_MAX_SIZE_HALF) + /* move the pointer to the correct range */ + *ptr += (SPI_TEST_MAX_SIZE_PLUS / 2) - + SPI_TEST_MAX_SIZE_HALF; + + /* RX range + * - we check against MAX_SIZE_PLUS to allow for automated alignment + */ + if (RANGE_CHECK(*ptr, len, RX(0), SPI_TEST_MAX_SIZE_PLUS)) { + off = *ptr - RX(0); + *ptr = rx + off; + + return 0; + } + + /* TX range */ + if (RANGE_CHECK(*ptr, len, TX(0), SPI_TEST_MAX_SIZE_PLUS)) { + off = *ptr - TX(0); + *ptr = tx + off; + + return 0; + } + + dev_err(&spi->dev, + "PointerRange [%pK:%pK[ not in range [%pK:%pK[ or [%pK:%pK[\n", + *ptr, *ptr + len, + RX(0), RX(SPI_TEST_MAX_SIZE), + TX(0), TX(SPI_TEST_MAX_SIZE)); + + return -EINVAL; +} + +static int spi_test_fill_pattern(struct spi_device *spi, + struct spi_test *test) +{ + struct spi_transfer *xfers = test->transfers; + u8 *tx_buf; + size_t count = 0; + int i, j; + +#ifdef __BIG_ENDIAN +#define GET_VALUE_BYTE(value, index, bytes) \ + (value >> (8 * (bytes - 1 - count % bytes))) +#else +#define GET_VALUE_BYTE(value, index, bytes) \ + (value >> (8 * (count % bytes))) +#endif + + /* fill all transfers with the pattern requested */ + for (i = 0; i < test->transfer_count; i++) { + /* fill rx_buf with SPI_TEST_PATTERN_UNWRITTEN */ + if (xfers[i].rx_buf) + memset(xfers[i].rx_buf, SPI_TEST_PATTERN_UNWRITTEN, + xfers[i].len); + /* if tx_buf is NULL then skip */ + tx_buf = (u8 *)xfers[i].tx_buf; + if (!tx_buf) + continue; + /* modify all the transfers */ + for (j = 0; j < xfers[i].len; j++, tx_buf++, count++) { + /* fill tx */ + switch (test->fill_option) { + case FILL_MEMSET_8: + *tx_buf = test->fill_pattern; + break; + case FILL_MEMSET_16: + *tx_buf = GET_VALUE_BYTE(test->fill_pattern, + count, 2); + break; + case FILL_MEMSET_24: + *tx_buf = GET_VALUE_BYTE(test->fill_pattern, + count, 3); + break; + case FILL_MEMSET_32: + *tx_buf = GET_VALUE_BYTE(test->fill_pattern, + count, 4); + break; + case FILL_COUNT_8: + *tx_buf = count; + break; + case FILL_COUNT_16: + *tx_buf = GET_VALUE_BYTE(count, count, 2); + break; + case FILL_COUNT_24: + *tx_buf = GET_VALUE_BYTE(count, count, 3); + break; + case FILL_COUNT_32: + *tx_buf = GET_VALUE_BYTE(count, count, 4); + break; + case FILL_TRANSFER_BYTE_8: + *tx_buf = j; + break; + case FILL_TRANSFER_BYTE_16: + *tx_buf = GET_VALUE_BYTE(j, j, 2); + break; + case FILL_TRANSFER_BYTE_24: + *tx_buf = GET_VALUE_BYTE(j, j, 3); + break; + case FILL_TRANSFER_BYTE_32: + *tx_buf = GET_VALUE_BYTE(j, j, 4); + break; + case FILL_TRANSFER_NUM: + *tx_buf = i; + break; + default: + dev_err(&spi->dev, + "unsupported fill_option: %i\n", + test->fill_option); + return -EINVAL; + } + } + } + + return 0; +} + +static int _spi_test_run_iter(struct spi_device *spi, + struct spi_test *test, + void *tx, void *rx) +{ + struct spi_message *msg = &test->msg; + struct spi_transfer *x; + int i, ret; + + /* initialize message - zero-filled via static initialization */ + spi_message_init_no_memset(msg); + + /* fill rx with the DO_NOT_WRITE pattern */ + memset(rx, SPI_TEST_PATTERN_DO_NOT_WRITE, SPI_TEST_MAX_SIZE_PLUS); + + /* add the individual transfers */ + for (i = 0; i < test->transfer_count; i++) { + x = &test->transfers[i]; + + /* patch the values of tx_buf */ + ret = spi_test_translate(spi, (void **)&x->tx_buf, x->len, + (void *)tx, rx); + if (ret) + return ret; + + /* patch the values of rx_buf */ + ret = spi_test_translate(spi, &x->rx_buf, x->len, + (void *)tx, rx); + if (ret) + return ret; + + /* and add it to the list */ + spi_message_add_tail(x, msg); + } + + /* fill in the transfer buffers with pattern */ + ret = spi_test_fill_pattern(spi, test); + if (ret) + return ret; + + /* and execute */ + if (test->execute_msg) + ret = test->execute_msg(spi, test, tx, rx); + else + ret = spi_test_execute_msg(spi, test, tx, rx); + + /* handle result */ + if (ret == test->expected_return) + return 0; + + dev_err(&spi->dev, + "test failed - test returned %i, but we expect %i\n", + ret, test->expected_return); + + if (ret) + return ret; + + /* if it is 0, as we expected something else, + * then return something special + */ + return -EFAULT; +} + +static int spi_test_run_iter(struct spi_device *spi, + const struct spi_test *testtemplate, + void *tx, void *rx, + size_t len, + size_t tx_off, + size_t rx_off + ) +{ + struct spi_test test; + int i, tx_count, rx_count; + + /* copy the test template to test */ + memcpy(&test, testtemplate, sizeof(test)); + + /* set up test->transfers to the correct count */ + if (!test.transfer_count) { + for (i = 0; + (i < SPI_TEST_MAX_TRANSFERS) && test.transfers[i].len; + i++) { + test.transfer_count++; + } + } + + /* if iterate_transfer_mask is not set, + * then set it to first transfer only + */ + if (!(test.iterate_transfer_mask & (BIT(test.transfer_count) - 1))) + test.iterate_transfer_mask = 1; + + /* count number of transfers with tx/rx_buf != NULL */ + rx_count = tx_count = 0; + for (i = 0; i < test.transfer_count; i++) { + if (test.transfers[i].tx_buf) + tx_count++; + if (test.transfers[i].rx_buf) + rx_count++; + } + + /* in some iteration cases warn and exit early, + * as there is nothing to do, that has not been tested already... + */ + if (tx_off && (!tx_count)) { + dev_warn_once(&spi->dev, + "%s: iterate_tx_off configured with tx_buf==NULL - ignoring\n", + test.description); + return 0; + } + if (rx_off && (!rx_count)) { + dev_warn_once(&spi->dev, + "%s: iterate_rx_off configured with rx_buf==NULL - ignoring\n", + test.description); + return 0; + } + + /* write out info */ + if (!(len || tx_off || rx_off)) { + dev_info(&spi->dev, "Running test %s\n", test.description); + } else { + dev_info(&spi->dev, + " with iteration values: len = %zu, tx_off = %zu, rx_off = %zu\n", + len, tx_off, rx_off); + } + + /* update in the values from iteration values */ + for (i = 0; i < test.transfer_count; i++) { + /* only when bit in transfer mask is set */ + if (!(test.iterate_transfer_mask & BIT(i))) + continue; + if (len) + test.transfers[i].len = len; + if (test.transfers[i].tx_buf) + test.transfers[i].tx_buf += tx_off; + if (test.transfers[i].tx_buf) + test.transfers[i].rx_buf += rx_off; + } + + /* and execute */ + return _spi_test_run_iter(spi, &test, tx, rx); +} + +/** + * spi_test_execute_msg - default implementation to run a test + * + * spi: @spi_device on which to run the @spi_message + * test: the test to execute, which already contains @msg + * tx: the tx buffer allocated for the test sequence + * rx: the rx buffer allocated for the test sequence + * + * Returns: error code of spi_sync as well as basic error checking + */ +int spi_test_execute_msg(struct spi_device *spi, struct spi_test *test, + void *tx, void *rx) +{ + struct spi_message *msg = &test->msg; + int ret = 0; + int i; + + /* only if we do not simulate */ + if (!simulate_only) { + /* dump the complete message before and after the transfer */ + if (dump_messages == 3) + spi_test_dump_message(spi, msg, true); + + /* run spi message */ + ret = spi_sync(spi, msg); + if (ret == -ETIMEDOUT) { + dev_info(&spi->dev, + "spi-message timed out - reruning...\n"); + /* rerun after a few explicit schedules */ + for (i = 0; i < 16; i++) + schedule(); + ret = spi_sync(spi, msg); + } + if (ret) { + dev_err(&spi->dev, + "Failed to execute spi_message: %i\n", + ret); + goto exit; + } + + /* do some extra error checks */ + if (msg->frame_length != msg->actual_length) { + dev_err(&spi->dev, + "actual length differs from expected\n"); + ret = -EIO; + goto exit; + } + + /* run rx-buffer tests */ + ret = spi_test_check_loopback_result(spi, msg, tx, rx); + } + + /* if requested or on error dump message (including data) */ +exit: + if (dump_messages || ret) + spi_test_dump_message(spi, msg, + (dump_messages >= 2) || (ret)); + + return ret; +} +EXPORT_SYMBOL_GPL(spi_test_execute_msg); + +/** + * spi_test_run_test - run an individual spi_test + * including all the relevant iterations on: + * length and buffer alignment + * + * spi: the spi_device to send the messages to + * test: the test which we need to execute + * tx: the tx buffer allocated for the test sequence + * rx: the rx buffer allocated for the test sequence + * + * Returns: status code of spi_sync or other failures + */ + +int spi_test_run_test(struct spi_device *spi, const struct spi_test *test, + void *tx, void *rx) +{ + int idx_len; + size_t len; + size_t tx_align, rx_align; + int ret; + + /* test for transfer limits */ + if (test->transfer_count >= SPI_TEST_MAX_TRANSFERS) { + dev_err(&spi->dev, + "%s: Exceeded max number of transfers with %i\n", + test->description, test->transfer_count); + return -E2BIG; + } + + /* setting up some values in spi_message + * based on some settings in spi_master + * some of this can also get done in the run() method + */ + + /* iterate over all the iterable values using macros + * (to make it a bit more readable... + */ +#define FOR_EACH_ITERATE(var, defaultvalue) \ + for (idx_##var = -1, var = defaultvalue; \ + ((idx_##var < 0) || \ + ( \ + (idx_##var < SPI_TEST_MAX_ITERATE) && \ + (var = test->iterate_##var[idx_##var]) \ + ) \ + ); \ + idx_##var++) +#define FOR_EACH_ALIGNMENT(var) \ + for (var = 0; \ + var < (test->iterate_##var ? \ + (spi->master->dma_alignment ? \ + spi->master->dma_alignment : \ + test->iterate_##var) : \ + 1); \ + var++) + + FOR_EACH_ITERATE(len, 0) { + FOR_EACH_ALIGNMENT(tx_align) { + FOR_EACH_ALIGNMENT(rx_align) { + /* and run the iteration */ + ret = spi_test_run_iter(spi, test, + tx, rx, + len, + tx_align, + rx_align); + if (ret) + return ret; + } + } + } + + return 0; +} +EXPORT_SYMBOL_GPL(spi_test_run_test); + +/** + * spi_test_run_tests - run an array of spi_messages tests + * @spi: the spi device on which to run the tests + * @tests: NULL-terminated array of @spi_test + * + * Returns: status errors as per @spi_test_run_test() + */ + +int spi_test_run_tests(struct spi_device *spi, + struct spi_test *tests) +{ + char *rx = NULL, *tx = NULL; + int ret = 0, count = 0; + struct spi_test *test; + + /* allocate rx/tx buffers of 128kB size without devm + * in the hope that is on a page boundary + */ + rx = kzalloc(SPI_TEST_MAX_SIZE_PLUS, GFP_KERNEL); + if (!rx) { + ret = -ENOMEM; + goto out; + } + + tx = kzalloc(SPI_TEST_MAX_SIZE_PLUS, GFP_KERNEL); + if (!tx) { + ret = -ENOMEM; + goto out; + } + + /* now run the individual tests in the table */ + for (test = tests, count = 0; test->description[0]; + test++, count++) { + /* only run test if requested */ + if ((run_only_test > -1) && (count != run_only_test)) + continue; + /* run custom implementation */ + if (test->run_test) + ret = test->run_test(spi, test, tx, rx); + else + ret = spi_test_run_test(spi, test, tx, rx); + if (ret) + goto out; + /* add some delays so that we can easily + * detect the individual tests when using a logic analyzer + * we also add scheduling to avoid potential spi_timeouts... + */ + mdelay(100); + schedule(); + } + +out: + kfree(rx); + kfree(tx); + return ret; +} +EXPORT_SYMBOL_GPL(spi_test_run_tests); -- cgit v1.2.3-54-g00ecf