linaro-kernel October 2013

linaro-kernel@lists.linaro.org

87 participants
222 discussions

[PATCH 0/2] additional vexpress-tc2 big endian fixes

by Victor Kamensky

Hi Ben, Here are couple more big endian related fixes. We run into these with vexpress TC2 board, when CONFIG_MCPM and CONFIG_ARM_CCI configs are enabled. Big endian image fails to boot. With fixes BE TC2 boots and it sees all 5 cores (2xA15, 3xA7). These were tested with vexpress TC2 on latest linux-linaro branch (include BE patch series) and it was tested on very recent rmk/fixes branch along with BE series on top of it. Thanks, Victor Victor Kamensky (2): ARM: mcpm: fix big endian issue in mcpm startup code ARM: cci: driver need big endian fixes in asm code arch/arm/common/mcpm_head.S | 2 ++ drivers/bus/arm-cci.c | 6 ++++++ 2 files changed, 8 insertions(+) -- 1.8.1.4

10 years, 7 months

[PATCH] clk: debugfs add clk_rate_fops with DEBUG

by Zhangfei Gao

clk_rate_fops is added to debug 1. set_rate (e.g. PLL) 2. Choose mux parent, since mux could choose parent accordingly when set_rate. Causion: set_rate can be called directly from user space Example: sfc_mux have two parents: 24M and 200M cat clk_summary clock enable_cnt prepare_cnt rate --------------------------------------------------------------------- osc24mhz 3 3 24000000 bpll 1 1 4294967295 bpll_fout3 0 0 200000000 sfc_mux 0 0 200000000 sfc 0 0 200000000 cat osc24mhz/bpll/bpll_fout3/sfc_mux/sfc/clk_rate 200000000 echo 24000000 > osc24mhz/bpll/bpll_fout3/sfc_mux/sfc/clk_rate cat clk_summary clock enable_cnt prepare_cnt rate --------------------------------------------------------------------- osc24mhz 3 3 24000000 sfc_mux 0 0 24000000 sfc 0 0 24000000 bpll 1 1 4294967295 bpll_fout3 0 0 200000000 cat osc24mhz/sfc_mux/sfc/clk_rate 24000000 Signed-off-by: Zhangfei Gao <zhangfei.gao(a)linaro.org> --- drivers/clk/clk.c | 34 ++++++++++++++++++++++++++++++++++ 1 file changed, 34 insertions(+) diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c index 373cd54..6b582f7 100644 --- a/drivers/clk/clk.c +++ b/drivers/clk/clk.c @@ -226,6 +226,35 @@ static const struct file_operations clk_dump_fops = { .release = single_release, }; +#ifdef DEBUG +static int clk_rate_fops_get(void *data, u64 *rate) +{ + struct clk *clk = data; + + *rate = clk->rate; + + return 0; +}; + +static int clk_rate_fops_set(void *data, u64 rate) +{ + struct clk *clk = data; + int ret = 0; + + ret = clk_prepare_enable(clk); + if (ret) + goto out; + clk_set_rate(clk, rate); + clk_disable_unprepare(clk); + +out: + return ret; +}; + +DEFINE_SIMPLE_ATTRIBUTE(clk_rate_fops, clk_rate_fops_get, + clk_rate_fops_set, "%llu\n"); +#endif + /* caller must hold prepare_lock */ static int clk_debug_create_one(struct clk *clk, struct dentry *pdentry) { @@ -243,8 +272,13 @@ static int clk_debug_create_one(struct clk *clk, struct dentry *pdentry) clk->dentry = d; +#ifdef DEBUG + d = debugfs_create_file("clk_rate", S_IWUSR | S_IRUGO, clk->dentry, + clk, &clk_rate_fops); +#else d = debugfs_create_u32("clk_rate", S_IRUGO, clk->dentry, (u32 *)&clk->rate); +#endif if (!d) goto err_out; -- 1.7.9.5

10 years, 7 months

linus/master boot: 33 pass, 0 fail (v3.12-rc4-19-g8b5ede6)

by Kevin Hilman build bot

Automated DT boot report for various ARM defconfigs. Boot test simply checks if kernel can boot to initramfs with busybox and run some basic commands (e.g. 'cat /proc/cpuinfo'). Tree/Branch: linus/master Git describe: v3.12-rc4-19-g8b5ede6 Commit: 8b5ede69d2 powerpc/irq: Don't switch to irq stack from softirq stack Full Report =========== arm-tegra_defconfig ------------------- tegra30-beaver PASS: 0 min 17.0 sec arm-omap2plus_defconfig ----------------------- am335x-boneblack PASS: 0 min 29.6 sec omap3-beagle-xm PASS: 0 min 48.9 sec am335x-bone PASS: 0 min 26.2 sec omap3-tobi,3530overo PASS: 0 min 21.3 sec omap4-panda PASS: 1 min 3.2 sec omap4-panda-es PASS: 1 min 5.3 sec omap3-tobi,3730storm PASS: 0 min 23.2 sec omap3-beagle PASS: 0 min 54.4 sec arm-imx_v6_v7_defconfig ----------------------- imx6dl-wandboard,wand-dual PASS: 0 min 16.1 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.0 sec imx6q-wandboard PASS: 0 min 14.8 sec arm-multi_v7_defconfig ---------------------- ste-snowball PASS: 0 min 32.2 sec tegra30-beaver PASS: 0 min 14.2 sec am335x-boneblack PASS: 0 min 34.4 sec omap3-beagle-xm PASS: 0 min 48.7 sec armada-370-mirabox PASS: 0 min 20.2 sec omap4-panda PASS: 0 min 59.8 sec omap4-panda-es PASS: 1 min 2.4 sec armada-xp-openblocks-ax3-4 PASS: 0 min 22.6 sec sun4i-a10-cubieboard PASS: 0 min 11.3 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.5 sec omap3-tobi,3530overo PASS: 0 min 19.5 sec am335x-bone PASS: 0 min 33.2 sec imx6q-wandboard PASS: 0 min 15.7 sec omap3-tobi,3730storm PASS: 0 min 22.2 sec imx6dl-wandboard,wand-dual PASS: 0 min 16.5 sec omap3-beagle PASS: 0 min 58.5 sec arm-u8500_defconfig ------------------- ste-snowball PASS: 0 min 41.0 sec arm-exynos_defconfig -------------------- exynos5250-arndale PASS: 0 min 29.7 sec arm-mvebu_defconfig ------------------- armada-xp-openblocks-ax3-4 PASS: 0 min 21.8 sec armada-370-mirabox PASS: 0 min 19.9 sec arm-sama5_defconfig ------------------- sama5d35ek PASS: 0 min 17.6 sec

10 years, 7 months

[PATCH] spi/trace: Trace length of SPI messages on completion

by Mark Brown

From: Mark Brown <broonie(a)linaro.org> Signed-off-by: Mark Brown <broonie(a)linaro.org> --- include/trace/events/spi.h | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) diff --git a/include/trace/events/spi.h b/include/trace/events/spi.h index a7b0907..8cd2d78 100644 --- a/include/trace/events/spi.h +++ b/include/trace/events/spi.h @@ -80,12 +80,30 @@ DEFINE_EVENT(spi_message, spi_message_start, ); -DEFINE_EVENT(spi_message, spi_message_done, +TRACE_EVENT(spi_message_done, TP_PROTO(struct spi_message *msg), - TP_ARGS(msg) + TP_ARGS(msg), + + TP_STRUCT__entry( + __field( int, bus_num ) + __field( int, chip_select ) + __field( struct spi_message *, msg ) + __field( unsigned, len ) + ), + TP_fast_assign( + __entry->bus_num = msg->spi->master->bus_num; + __entry->chip_select = msg->spi->chip_select; + __entry->msg = msg; + __entry->len = msg->actual_length; + ), + + TP_printk("spi%d.%d %p len=%u", (int)__entry->bus_num, + (int)__entry->chip_select, + (struct spi_message *)__entry->msg, + (unsigned)__entry->len) ); #endif /* _TRACE_POWER_H */ -- 1.8.4.rc3

10 years, 7 months

arm-soc/to-build boot: 29 pass, 4 fail (v3.12-rc2-281-ge18181c)

by Kevin Hilman build bot

Automated DT boot report for various ARM defconfigs. Boot test simply checks if kernel can boot to initramfs with busybox and run some basic commands (e.g. 'cat /proc/cpuinfo'). Tree/Branch: arm-soc/to-build Git describe: v3.12-rc2-281-ge18181c Commit: e18181c565 Merge branch 'next/cleanup' into for-next Failed boot tests (console logs at the end) =========================================== ste-snowball: FAIL: arm-multi_v7_defconfig ste-snowball: FAIL: arm-u8500_defconfig armada-xp-openblocks-ax3-4: FAIL: arm-mvebu_defconfig armada-370-mirabox: FAIL: arm-mvebu_defconfig Full Report =========== arm-tegra_defconfig ------------------- tegra30-beaver PASS: 0 min 16.9 sec arm-omap2plus_defconfig ----------------------- am335x-boneblack PASS: 0 min 29.9 sec omap3-beagle-xm PASS: 0 min 51.9 sec am335x-bone PASS: 0 min 26.1 sec omap3-tobi,3530overo PASS: 0 min 42.4 sec omap4-panda PASS: 1 min 6.3 sec omap4-panda-es PASS: 1 min 5.4 sec omap3-tobi,3730storm PASS: 0 min 41.4 sec omap3-beagle PASS: 0 min 52.9 sec arm-imx_v6_v7_defconfig ----------------------- imx6dl-wandboard,wand-dual PASS: 0 min 16.3 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.2 sec imx6q-wandboard PASS: 0 min 14.8 sec arm-multi_v7_defconfig ---------------------- ste-snowball FAIL: 0 min 33.8 sec tegra30-beaver PASS: 0 min 14.2 sec am335x-boneblack PASS: 0 min 34.4 sec omap3-beagle-xm PASS: 0 min 51.1 sec armada-370-mirabox PASS: 0 min 20.4 sec omap4-panda PASS: 1 min 1.1 sec omap4-panda-es PASS: 1 min 8.8 sec armada-xp-openblocks-ax3-4 PASS: 0 min 22.9 sec sun4i-a10-cubieboard PASS: 0 min 11.3 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.3 sec omap3-tobi,3530overo PASS: 0 min 40.6 sec am335x-bone PASS: 0 min 33.2 sec imx6q-wandboard PASS: 0 min 15.7 sec omap3-tobi,3730storm PASS: 0 min 43.2 sec imx6dl-wandboard,wand-dual PASS: 0 min 16.6 sec omap3-beagle PASS: 0 min 48.8 sec arm-u8500_defconfig ------------------- ste-snowball FAIL: 0 min 30.0 sec arm-exynos_defconfig -------------------- exynos5250-arndale PASS: 0 min 28.9 sec arm-mvebu_defconfig ------------------- armada-xp-openblocks-ax3-4 FAIL: 0 min 28.8 sec armada-370-mirabox FAIL: 0 min 25.9 sec arm-sama5_defconfig ------------------- sama5d35ek PASS: 0 min 18.0 sec Console logs for failures ========================= arm-multi_v7_defconfig ---------------------- ste-snowball: FAIL: last 24 lines of boot log: ---------------------------------------------- Load address: 0x1000000 Loading: *## done Bytes transferred = 18596 (48a4 hex) U8500 $ printenv bootargs printenv bootargs bootargs=console=ttyAMA2,115200n8 debug earlyprintk rw root=/dev/mmcblk0p3 rootwait rootfstype=ext4 ip=192.168.1.184:192.168.1.2:192.168.1.254:255.255.255.0::::192.168.1.254:none U8500 $ bootm 0x00800000 - 0x01000000 bootm 0x00800000 - 0x01000000 ## Booting kernel from Legacy Image at 00800000 ... Image Name: Linux Image Type: ARM Linux Kernel Image (uncompressed) Data Size: 3576784 Bytes = 3.4 MB Load Address: 00008000 Entry Point: 00008000 Loading Kernel Image ... OK OK Starting kernel ... ~$off # PYBOOT: Exception: kernel: ERROR: did not start booting. # PYBOOT: Time: 33.78 seconds. # PYBOOT: Result: FAIL arm-u8500_defconfig ------------------- ste-snowball: FAIL: last 24 lines of boot log: ---------------------------------------------- Load address: 0x1000000 Loading: *## done Bytes transferred = 18596 (48a4 hex) U8500 $ printenv bootargs printenv bootargs bootargs=console=ttyAMA2,115200n8 debug earlyprintk rw root=/dev/mmcblk0p3 rootwait rootfstype=ext4 ip=192.168.1.184:192.168.1.2:192.168.1.254:255.255.255.0::::192.168.1.254:none U8500 $ bootm 0x00800000 - 0x01000000 bootm 0x00800000 - 0x01000000 ## Booting kernel from Legacy Image at 00800000 ... Image Name: Linux Image Type: ARM Linux Kernel Image (uncompressed) Data Size: 2846208 Bytes = 2.7 MB Load Address: 00008000 Entry Point: 00008000 Loading Kernel Image ... OK OK Starting kernel ... ~$off # PYBOOT: Exception: kernel: ERROR: did not start booting. # PYBOOT: Time: 30.01 seconds. # PYBOOT: Result: FAIL arm-mvebu_defconfig ------------------- armada-xp-openblocks-ax3-4: FAIL: last 24 lines of boot log: ------------------------------------------------------------ Image Name: Linux Created: 2013-10-07 22:10:49 UTC Image Type: ARM Linux Kernel Image (uncompressed) Data Size: 2659644 Bytes = 2.5 MiB Load Address: 00008000 Entry Point: 00008000 Verifying Checksum ... OK ## Loading init Ramdisk from Legacy Image at 03000000 ... Image Name: Created: 2013-09-07 5:02:12 UTC Image Type: ARM Linux RAMDisk Image (uncompressed) Data Size: 639744 Bytes = 624.8 KiB Load Address: 81000000 Entry Point: 81000000 Verifying Checksum ... OK Loading Kernel Image ... OK OK Starting kernel ... ~$off # PYBOOT: Exception: kernel: ERROR: did not start booting. # PYBOOT: Time: 28.79 seconds. # PYBOOT: Result: FAIL armada-370-mirabox: FAIL: last 24 lines of boot log: ---------------------------------------------------- Image Name: Linux Created: 2013-10-07 22:10:23 UTC Image Type: ARM Linux Kernel Image (uncompressed) Data Size: 2656025 Bytes = 2.5 MB Load Address: 00008000 Entry Point: 00008000 Verifying Checksum ... OK ## Loading init Ramdisk from Legacy Image at 03000000 ... Image Name: Created: 2013-09-07 5:02:12 UTC Image Type: ARM Linux RAMDisk Image (uncompressed) Data Size: 639744 Bytes = 624.8 kB Load Address: 81000000 Entry Point: 81000000 Verifying Checksum ... OK Loading Kernel Image ... OK OK Starting kernel ... ~$off # PYBOOT: Exception: kernel: ERROR: did not start booting. # PYBOOT: Time: 25.93 seconds. # PYBOOT: Result: FAIL

10 years, 7 months

[PATCH] regmap: Cache async work structures

by Mark Brown

From: Mark Brown <broonie(a)linaro.org> Rather than allocating and deallocating the structures used to manage async transfers each time we do one keep the structures around as long as the regmap is around. This should provide a small performance improvement. Signed-off-by: Mark Brown <broonie(a)linaro.org> --- drivers/base/regmap/internal.h | 2 +- drivers/base/regmap/regmap.c | 59 +++++++++++++++++++++++------------------- 2 files changed, 34 insertions(+), 27 deletions(-) diff --git a/drivers/base/regmap/internal.h b/drivers/base/regmap/internal.h index 9010614..63c22bc 100644 --- a/drivers/base/regmap/internal.h +++ b/drivers/base/regmap/internal.h @@ -44,7 +44,6 @@ struct regmap_format { struct regmap_async { struct list_head list; - struct work_struct cleanup; struct regmap *map; void *work_buf; }; @@ -67,6 +66,7 @@ struct regmap { spinlock_t async_lock; wait_queue_head_t async_waitq; struct list_head async_list; + struct list_head async_free; int async_ret; #ifdef CONFIG_DEBUG_FS diff --git a/drivers/base/regmap/regmap.c b/drivers/base/regmap/regmap.c index 00152bf..5754513 100644 --- a/drivers/base/regmap/regmap.c +++ b/drivers/base/regmap/regmap.c @@ -42,15 +42,6 @@ static int _regmap_bus_formatted_write(void *context, unsigned int reg, static int _regmap_bus_raw_write(void *context, unsigned int reg, unsigned int val); -static void async_cleanup(struct work_struct *work) -{ - struct regmap_async *async = container_of(work, struct regmap_async, - cleanup); - - kfree(async->work_buf); - kfree(async); -} - bool regmap_reg_in_ranges(unsigned int reg, const struct regmap_range *ranges, unsigned int nranges) @@ -465,6 +456,7 @@ struct regmap *regmap_init(struct device *dev, spin_lock_init(&map->async_lock); INIT_LIST_HEAD(&map->async_list); + INIT_LIST_HEAD(&map->async_free); init_waitqueue_head(&map->async_waitq); if (config->read_flag_mask || config->write_flag_mask) { @@ -944,12 +936,22 @@ EXPORT_SYMBOL_GPL(regmap_reinit_cache); */ void regmap_exit(struct regmap *map) { + struct regmap_async *async; + regcache_exit(map); regmap_debugfs_exit(map); regmap_range_exit(map); if (map->bus && map->bus->free_context) map->bus->free_context(map->bus_context); kfree(map->work_buf); + while (!list_empty(&map->async_free)) { + async = list_first_entry_or_null(&map->async_free, + struct regmap_async, + list); + list_del(&async->list); + kfree(async->work_buf); + kfree(async); + } kfree(map); } EXPORT_SYMBOL_GPL(regmap_exit); @@ -1117,20 +1119,31 @@ int _regmap_raw_write(struct regmap *map, unsigned int reg, u8[0] |= map->write_flag_mask; if (async && map->bus->async_write) { - struct regmap_async *async = map->bus->async_alloc(); - if (!async) - return -ENOMEM; + struct regmap_async *async; trace_regmap_async_write_start(map->dev, reg, val_len); - async->work_buf = kzalloc(map->format.buf_size, - GFP_KERNEL | GFP_DMA); - if (!async->work_buf) { - kfree(async); - return -ENOMEM; + spin_lock_irqsave(&map->async_lock, flags); + async = list_first_entry_or_null(&map->async_free, + struct regmap_async, + list); + if (async) + list_del(&async->list); + spin_unlock_irqrestore(&map->async_lock, flags); + + if (!async) { + async = map->bus->async_alloc(); + if (!async) + return -ENOMEM; + + async->work_buf = kzalloc(map->format.buf_size, + GFP_KERNEL | GFP_DMA); + if (!async->work_buf) { + kfree(async); + return -ENOMEM; + } } - INIT_WORK(&async->cleanup, async_cleanup); async->map = map; /* If the caller supplied the value we can use it safely. */ @@ -1154,11 +1167,8 @@ int _regmap_raw_write(struct regmap *map, unsigned int reg, ret); spin_lock_irqsave(&map->async_lock, flags); - list_del(&async->list); + list_move(&async->list, &map->async_free); spin_unlock_irqrestore(&map->async_lock, flags); - - kfree(async->work_buf); - kfree(async); } return ret; @@ -1923,8 +1933,7 @@ void regmap_async_complete_cb(struct regmap_async *async, int ret) trace_regmap_async_io_complete(map->dev); spin_lock(&map->async_lock); - - list_del(&async->list); + list_move(&async->list, &map->async_free); wake = list_empty(&map->async_list); if (ret != 0) @@ -1932,8 +1941,6 @@ void regmap_async_complete_cb(struct regmap_async *async, int ret) spin_unlock(&map->async_lock); - schedule_work(&async->cleanup); - if (wake) wake_up(&map->async_waitq); } -- 1.8.4.rc3

10 years, 7 months

linus/master boot: 32 pass, 1 fail (v3.12-rc4-18-gfd84831)

by Kevin Hilman build bot

Automated DT boot report for various ARM defconfigs. Boot test simply checks if kernel can boot to initramfs with busybox and run some basic commands (e.g. 'cat /proc/cpuinfo'). Tree/Branch: linus/master Git describe: v3.12-rc4-18-gfd84831 Commit: fd848319e7 Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid Failed boot tests (console logs at the end) =========================================== omap4-panda: FAIL: arm-multi_v7_defconfig Full Report =========== arm-tegra_defconfig ------------------- tegra30-beaver PASS: 0 min 17.1 sec arm-omap2plus_defconfig ----------------------- am335x-boneblack PASS: 0 min 29.8 sec omap3-beagle-xm PASS: 0 min 51.8 sec am335x-bone PASS: 0 min 26.2 sec omap3-tobi,3530overo PASS: 0 min 24.4 sec omap4-panda PASS: 1 min 8.8 sec omap4-panda-es PASS: 1 min 7.7 sec omap3-tobi,3730storm PASS: 0 min 23.3 sec omap3-beagle PASS: 0 min 56.1 sec arm-imx_v6_v7_defconfig ----------------------- imx6dl-wandboard,wand-dual PASS: 0 min 16.1 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.1 sec imx6q-wandboard PASS: 0 min 14.9 sec arm-multi_v7_defconfig ---------------------- ste-snowball PASS: 0 min 30.3 sec tegra30-beaver PASS: 0 min 16.6 sec am335x-boneblack PASS: 0 min 34.4 sec omap3-beagle-xm PASS: 0 min 52.4 sec armada-370-mirabox PASS: 0 min 20.4 sec omap4-panda FAIL: 2 min 55.2 sec omap4-panda-es PASS: 1 min 3.9 sec armada-xp-openblocks-ax3-4 PASS: 0 min 22.6 sec sun4i-a10-cubieboard PASS: 0 min 11.3 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.5 sec omap3-tobi,3530overo PASS: 0 min 19.7 sec am335x-bone PASS: 0 min 33.3 sec imx6q-wandboard PASS: 0 min 15.8 sec omap3-tobi,3730storm PASS: 0 min 22.6 sec imx6dl-wandboard,wand-dual PASS: 0 min 16.6 sec omap3-beagle PASS: 0 min 49.3 sec arm-u8500_defconfig ------------------- ste-snowball PASS: 0 min 35.6 sec arm-exynos_defconfig -------------------- exynos5250-arndale PASS: 0 min 29.8 sec arm-mvebu_defconfig ------------------- armada-xp-openblocks-ax3-4 PASS: 0 min 21.9 sec armada-370-mirabox PASS: 0 min 19.9 sec arm-sama5_defconfig ------------------- sama5d35ek PASS: 0 min 18.4 sec Console logs for failures ========================= arm-multi_v7_defconfig ---------------------- omap4-panda: FAIL: last 24 lines of boot log: --------------------------------------------- [ 1.612457] omap_hsmmc 4809c000.mmc: pins are not configured from the driver [ 1.657348] pinctrl-single 4a31e040.pinmux: mux offset out of range: 0x38 (0x38) [ 1.665435] pinctrl-single 4a31e040.pinmux: could not add functions for pinmux_wl12xx_pins 56x [ 1.674682] omap-dma-engine 4a056000.dma-controller: allocating channel for 60 [ 1.682434] omap-dma-engine 4a056000.dma-controller: allocating channel for 59 [ 1.707275] usb 1-1: new high-speed USB device number 2 using ehci-omap [ 1.842071] twl_rtc rtc.11: setting system clock to 2013-10-07 19:19:56 UTC (1381173596) [ 1.858123] hub 1-1:1.0: USB hub found [ 1.862396] hub 1-1:1.0: 5 ports detected [ 1.939300] mmc0: host does not support reading read-only switch. assuming write-enable. [ 1.949981] mmc0: new high speed SD card at address aaaa [ 1.956054] mmcblk0: mmc0:aaaa SD02G 1.84 GiB [ 1.962432] mmcblk0: p1 p2 [ 2.157379] usb 1-1.1: new high-speed USB device number 3 using ehci-omap [ 16.050537] EXT4-fs (mmcblk0p2): recovery complete [ 16.063140] EXT4-fs (mmcblk0p2): mounted filesystem with ordered data mode. Opts: (null) [ 16.071777] VFS: Mounted root (ext4 filesystem) on device 179:2. [ 16.079345] devtmpfs: mounted [ 16.079467] Freeing unused kernel memory: 464K (c0629000 - c069d000) [ 16.495941] init: ureadahead main process (56) terminated with status 5 ~$off # PYBOOT: Exception: userspace: ERROR: Never found root shell # PYBOOT: Time: 175.23 seconds. # PYBOOT: Result: FAIL

10 years, 7 months

[PATCH] spi: Provide trace points for message processing

by Mark Brown

From: Mark Brown <broonie(a)linaro.org> Provide tracepoints for the lifecycle of a message from submission to completion and for the active time for masters to help with performance analysis of SPI I/O. Signed-off-by: Mark Brown <broonie(a)linaro.org> --- drivers/spi/spi.c | 13 +++++++ include/trace/events/spi.h | 94 ++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 107 insertions(+) create mode 100644 include/trace/events/spi.h diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c index 001cdc4..29ef371 100644 --- a/drivers/spi/spi.c +++ b/drivers/spi/spi.c @@ -39,6 +39,9 @@ #include <linux/ioport.h> #include <linux/acpi.h> +#define CREATE_TRACE_POINTS +#include <trace/events/spi.h> + static void spidev_release(struct device *dev) { struct spi_device *spi = to_spi_device(dev); @@ -557,6 +560,7 @@ static void spi_pump_messages(struct kthread_work *work) pm_runtime_mark_last_busy(master->dev.parent); pm_runtime_put_autosuspend(master->dev.parent); } + trace_spi_master_idle(master); return; } @@ -585,6 +589,9 @@ static void spi_pump_messages(struct kthread_work *work) } } + if (!was_busy) + trace_spi_master_busy(master); + if (!was_busy && master->prepare_transfer_hardware) { ret = master->prepare_transfer_hardware(master); if (ret) { @@ -597,6 +604,8 @@ static void spi_pump_messages(struct kthread_work *work) } } + trace_spi_message_start(master->cur_msg); + ret = master->transfer_one_message(master, master->cur_msg); if (ret) { dev_err(&master->dev, @@ -689,6 +698,8 @@ void spi_finalize_current_message(struct spi_master *master) mesg->state = NULL; if (mesg->complete) mesg->complete(mesg->context); + + trace_spi_message_done(mesg); } EXPORT_SYMBOL_GPL(spi_finalize_current_message); @@ -1450,6 +1461,8 @@ static int __spi_async(struct spi_device *spi, struct spi_message *message) struct spi_master *master = spi->master; struct spi_transfer *xfer; + trace_spi_message_submit(message); + if (list_empty(&message->transfers)) return -EINVAL; if (!message->complete) diff --git a/include/trace/events/spi.h b/include/trace/events/spi.h new file mode 100644 index 0000000..a7b0907 --- /dev/null +++ b/include/trace/events/spi.h @@ -0,0 +1,94 @@ +#undef TRACE_SYSTEM +#define TRACE_SYSTEM spi + +#if !defined(_TRACE_SPI_H) || defined(TRACE_HEADER_MULTI_READ) +#define _TRACE_SPI_H + +#include <linux/ktime.h> +#include <linux/tracepoint.h> + +DECLARE_EVENT_CLASS(spi_master, + + TP_PROTO(struct spi_master *master), + + TP_ARGS(master), + + TP_STRUCT__entry( + __field( int, bus_num ) + ), + + TP_fast_assign( + __entry->bus_num = master->bus_num; + ), + + TP_printk("spi%d", (int)__entry->bus_num) + +); + +DEFINE_EVENT(spi_master, spi_master_idle, + + TP_PROTO(struct spi_master *master), + + TP_ARGS(master) + +); + +DEFINE_EVENT(spi_master, spi_master_busy, + + TP_PROTO(struct spi_master *master), + + TP_ARGS(master) + +); + +DECLARE_EVENT_CLASS(spi_message, + + TP_PROTO(struct spi_message *msg), + + TP_ARGS(msg), + + TP_STRUCT__entry( + __field( int, bus_num ) + __field( int, chip_select ) + __field( struct spi_message *, msg ) + ), + + TP_fast_assign( + __entry->bus_num = msg->spi->master->bus_num; + __entry->chip_select = msg->spi->chip_select; + __entry->msg = msg; + ), + + TP_printk("spi%d.%d %p", (int)__entry->bus_num, + (int)__entry->chip_select, + (struct spi_message *)__entry->msg) +); + +DEFINE_EVENT(spi_message, spi_message_submit, + + TP_PROTO(struct spi_message *msg), + + TP_ARGS(msg) + +); + +DEFINE_EVENT(spi_message, spi_message_start, + + TP_PROTO(struct spi_message *msg), + + TP_ARGS(msg) + +); + +DEFINE_EVENT(spi_message, spi_message_done, + + TP_PROTO(struct spi_message *msg), + + TP_ARGS(msg) + +); + +#endif /* _TRACE_POWER_H */ + +/* This part must be outside protection */ +#include <trace/define_trace.h> -- 1.8.4.rc3

10 years, 7 months

linus/HEAD boot: 33 pass, 0 fail (v3.12-rc4-6-g162bdaf)

by Kevin Hilman build bot

Automated DT boot report for various ARM defconfigs. Boot test simply checks if kernel can boot to initramfs with busybox and run some basic commands (e.g. 'cat /proc/cpuinfo'). Tree/Branch: linus/HEAD Git describe: v3.12-rc4-6-g162bdaf Commit: 162bdafa46 Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux Full Report =========== arm-tegra_defconfig ------------------- tegra30-beaver PASS: 0 min 16.9 sec arm-omap2plus_defconfig ----------------------- am335x-boneblack PASS: 0 min 34.7 sec omap3-beagle-xm PASS: 0 min 50.8 sec am335x-bone PASS: 0 min 26.1 sec omap3-tobi,3530overo PASS: 0 min 24.3 sec omap4-panda PASS: 1 min 11.3 sec omap4-panda-es PASS: 1 min 10.9 sec omap3-tobi,3730storm PASS: 0 min 23.2 sec omap3-beagle PASS: 0 min 50.3 sec arm-imx_v6_v7_defconfig ----------------------- imx6dl-wandboard,wand-dual PASS: 0 min 16.2 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.2 sec imx6q-wandboard PASS: 0 min 14.9 sec arm-multi_v7_defconfig ---------------------- ste-snowball PASS: 0 min 30.6 sec tegra30-beaver PASS: 0 min 16.9 sec am335x-boneblack PASS: 0 min 39.4 sec omap3-beagle-xm PASS: 0 min 52.2 sec armada-370-mirabox PASS: 0 min 20.1 sec omap4-panda PASS: 1 min 0.3 sec omap4-panda-es PASS: 1 min 10.3 sec armada-xp-openblocks-ax3-4 PASS: 0 min 22.6 sec sun4i-a10-cubieboard PASS: 0 min 14.3 sec imx6dl-wandboard,wand-solo PASS: 0 min 16.4 sec omap3-tobi,3530overo PASS: 0 min 19.6 sec am335x-bone PASS: 0 min 33.2 sec imx6q-wandboard PASS: 0 min 15.8 sec omap3-tobi,3730storm PASS: 0 min 22.2 sec imx6dl-wandboard,wand-dual PASS: 0 min 16.6 sec omap3-beagle PASS: 0 min 55.0 sec arm-u8500_defconfig ------------------- ste-snowball PASS: 0 min 38.6 sec arm-exynos_defconfig -------------------- exynos5250-arndale PASS: 0 min 29.0 sec arm-mvebu_defconfig ------------------- armada-xp-openblocks-ax3-4 PASS: 0 min 22.2 sec armada-370-mirabox PASS: 0 min 19.8 sec arm-sama5_defconfig ------------------- sama5d35ek PASS: 0 min 18.4 sec

10 years, 7 months

[PATCH] ARM: tlb: __flush_tlb_mm need to use int asid var for BE correct operation

by Victor Kamensky

Hi Will, Could you please review patch following this cover letter. I've run into this problem when Linaro big endian topic branch received 3.12-rc2 update and as result got your f0915781bd5edf78b1154e61efe962dc15872d09 "ARM: tlb: don't perform inner-shareable invalidation for local TLB ops" commit. My Big Endian Arndale image was getting very weird user-land crashes while fork glib call was performed. It turns out that in __flush_tlb_mm function code passes 'ASID(mm)' directly to tlb_op macro, unlike in other functions (i.e local_flush_tlb_mm) where asid initially assigned to local variable with 'int' type and then such variable is passed to tlb_op macro. Direct use of 'ASID(mm)' in tlb_op macro does not work in big endian case. Resulting generated code look like this (from flush_tlb_mm that uses __flush_tlb_mm): (gdb) disassemble flush_tlb_mm Dump of assembler code for function flush_tlb_mm: <snip> 0x80011984 <+32>: dsb ishst 0x80011988 <+36>: movs r4, #0 <--------------------- 0x8001198a <+38>: ldrb.w r5, [r2, #375] ; 0x177 0x8001198e <+42>: ldr r3, [r3, #8] 0x80011990 <+44>: tst.w r3, #65536 ; 0x10000 0x80011994 <+48>: it ne 0x80011996 <+50>: mcrne 15, 0, r5, cr8, cr7, {2} 0x8001199a <+54>: tst.w r1, #4194304 ; 0x400000 0x8001199e <+58>: it ne 0x800119a0 <+60>: mcrne 15, 0, r4, cr8, cr3, {2} <------- 0x800119a4 <+64>: dsb ish '15, 0, r4, cr8, cr3, {2}' receives 0 through r4 register. Note that in LE case correct code is generated. If I change code to use intermediate int variable asid as other places do, correct code is generated for __flush_tlb_mm. My explanation is that ASID macro actually produces 64 bit integer. When passed to inline assembler as 'r' operand is not really defined behaviour and it is wrong in BE case. When intermidiate 'int' variable is used ASID macro value is properly converted to 32 bit integer. And inline assembler received int type, which works correctly. Note other possible ways to fix it is to add 'int' cast either outside of ASID macro call or inside of it. Personally I prefer variant that follows this cover letter, because it is similar to other cases where such code pattern is used. I've tried to understand whether proposed fix is really workaround for compiler bug. I failed to find clear explanation how gcc need to handle such situation. Here is quote from gcc manual: > The compiler cannot > check whether the operands have data types that are reasonable for the > instruction being executed. It does not parse the assembler instruction > template and does not know what it means or even whether it is valid > assembler input. The extended `asm' feature is most often used for > machine instructions the compiler itself does not know exist. If the > output expression cannot be directly addressed (for example, it is a > bit-field), your constraint must allow a register. In that case, GCC > will use the register as the output of the `asm', and then store that > register into the output. It is not clear whether 'unsigned long long' is "reasonable" type for arm "r" inline asm operand. So it seems it is gray area. Personally I suspect this issue came up before because in many places of tlbflush.h 'zero' local variable is used and 'asid' local variable is used. Here is small test case that illustrate code generation issue and difference between LE and BE cases: [kamensky@kamensky-w530 asid_inline]$ cat asid_inline.c typedef unsigned long long u64; struct mm_struct { unsigned long dummy1; u64 __attribute__((aligned(8))) counter; }; void test1(struct mm_struct *mm) { const int asid = ((mm)->counter & ~((~0ULL) << 8)); do { asm("mcr " "p15, 0, %0, " "c8, c3, 2" : : "r" (asid) : "cc"); } while (0); } void test2(struct mm_struct *mm) { do { asm("mcr " "p15, 0, %0, " "c8, c3, 2" : : "r" (((mm)->counter & ~((~0ULL) << 8))) : "cc"); } while (0); } void test3(struct mm_struct *mm) { do { asm("mcr " "p15, 0, %0, " "c8, c3, 2" : : "r" ((int)((mm)->counter & ~((~0ULL) << 8))) : "cc"); } while (0); } [kamensky@kamensky-w530 asid_inline]$ ./asid_inline.sh + arm-linux-gnueabihf-gcc -nostdinc -mbig-endian -O2 -mabi=aapcs-linux -mno-thumb-interwork -mthumb -march=armv7-a -msoft-float -c -o asid_inline.be.o asid_inline.c + arm-linux-gnueabihf-objdump --disassemble --reloc asid_inline.be.o asid_inline.be.o: file format elf32-bigarm Disassembly of section .text: 00000000 <test1>: 0: 7bc3 ldrb r3, [r0, #15] 2: ee08 3f53 mcr 15, 0, r3, cr8, cr3, {2} 6: 4770 bx lr 00000008 <test2>: 8: 7bc3 ldrb r3, [r0, #15] a: 2200 movs r2, #0 c: ee08 2f53 mcr 15, 0, r2, cr8, cr3, {2} 10: 4770 bx lr 12: bf00 nop 00000014 <test3>: 14: 7bc3 ldrb r3, [r0, #15] 16: ee08 3f53 mcr 15, 0, r3, cr8, cr3, {2} 1a: 4770 bx lr + arm-linux-gnueabihf-gcc -nostdinc -mlittle-endian -O2 -mabi=aapcs-linux -mno-thumb-interwork -mthumb -march=armv7-a -msoft-float -c -o asid_inline.le.o asid_inline.c + arm-linux-gnueabihf-objdump --disassemble --reloc asid_inline.le.o asid_inline.le.o: file format elf32-littlearm Disassembly of section .text: 00000000 <test1>: 0: 7a03 ldrb r3, [r0, #8] 2: ee08 3f53 mcr 15, 0, r3, cr8, cr3, {2} 6: 4770 bx lr 00000008 <test2>: 8: 7a02 ldrb r2, [r0, #8] a: 2300 movs r3, #0 c: ee08 2f53 mcr 15, 0, r2, cr8, cr3, {2} 10: 4770 bx lr 12: bf00 nop 00000014 <test3>: 14: 7a03 ldrb r3, [r0, #8] 16: ee08 3f53 mcr 15, 0, r3, cr8, cr3, {2} 1a: 4770 bx lr [kamensky@kamensky-w530 asid_inline]$ Thanks, Victor Victor Kamensky (1): ARM: tlb: __flush_tlb_mm need to use int asid var for BE correct operation arch/arm/include/asm/tlbflush.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) -- 1.8.1.4

10 years, 7 months

← Newer
1
...
16
17
18
19
20
21
22
23
Older →

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

linaro-kernel October 2013