drm/trace: Mirror DRM debug logs to tracefs
Subsystems (10)
Sean Paul
reviewerv5 cover drm/trace: Mirror DRM debug logs to tracefs
From: Sean Paul <[email protected]>
This series is the latest in my journey to create a lightweight, always-on "flight recorder" (name credit Weston) of drm logs. This incarnation uses a trace_array to keep logs in memory exposed through tracefs. Users and distros can enable drm logs by using the drm.trace module parameter to choose the debug categories they are interested in.
The set has ballooned a little bit since the last version. Reason being is that I decided to return true from drm_debug_enabled if trace was enabled. This should make it easier and more seamless for driver developers to use the correct interface, but meant I needed to audit all uses of drm_debug_enabled and drm_debug_printer.
Out of all those calls, there are 3 situations which arose:
1- The logs should only go to syslog:
I've converted these to use the drm_debug_syslog variant of enable check/printer.
2- The logs should go to both syslog/trace, but were using pr_debug():
I've converted these to use the proper drm logging helper. In cases which used a drm_printer I've had to use a new drm_debug_category_printer to ensure they are routed correctly.
3- The logs should go to both syslog/trace and are using drm logging fns:
These are untouched and should be routed to the appropriate place(s)
Sean Paul (13):
drm/mipi_dbi: Convert pr_debug calls to DRM_DEBUG_DRIVER drm/sil164: Convert dev_printk to drm_dev_dbg drm/i915/utils: Replace dev_printk with drm helpers drm/msm/dpu: Replace definitions for dpu debug macros drm/print: rename drm_debug* to be more syslog-centric drm/amd: Gate i2c transaction logs on drm_debug_syslog drm/etnaviv: Change buffer dump checks to target syslog drm/nouveau: Change debug checks to specifically target syslog drm/i915: Change infoframe debug checks to specify syslog drm/print: Add drm_debug_category_printer
drm/mst: Convert debug printers to debug category printers drm/i915: Use debug category printer for welcome message drm/print: Add tracefs support to the drm logging helpers
Documentation/gpu/drm-uapi.rst | 6 + drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c | 4 +- drivers/gpu/drm/drm_dp_mst_topology.c | 9 +- drivers/gpu/drm/drm_drv.c | 3 + drivers/gpu/drm/drm_mipi_dbi.c | 8 +- drivers/gpu/drm/drm_print.c | 228 ++++++++++++++++--- drivers/gpu/drm/etnaviv/etnaviv_buffer.c | 8 +- drivers/gpu/drm/i2c/sil164_drv.c | 12 +- drivers/gpu/drm/i915/display/intel_display.c | 4 +- drivers/gpu/drm/i915/i915_drv.c | 3 +- drivers/gpu/drm/i915/i915_utils.c | 5 +- drivers/gpu/drm/msm/disp/dpu1/dpu_kms.h | 20 +- drivers/gpu/drm/nouveau/dispnv50/disp.h | 4 +- drivers/gpu/drm/nouveau/nouveau_drv.h | 4 +- include/drm/drm_print.h | 96 +++++++- 15 files changed, 318 insertions(+), 96 deletions(-)
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 1/13 drm/mipi_dbi: Convert pr_debug calls to DRM_DEBUG_DRIVER
From: Sean Paul <[email protected]>
Use the drm logging helpers to output these messages to ensure they'll be included by the drm tracefs instance.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/drm_mipi_dbi.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-)
drivers/gpu/drm/drm_mipi_dbi.c +2 -6
@@ -769,9 +769,7 @@ static int mipi_dbi_spi1e_transfer(struct mipi_dbi *dbi, int dc,int i, ret;u8 *dst; -if (drm_debug_enabled (DRM_UT_DRIVER )) -pr_debug ("[drm:%s] dc=%d, max_chunk=%zu, transfers:\n" , - __func__, dc, max_chunk); +DRM_DEBUG_DRIVER ("dc=%d, max_chunk=%zu, transfers:\n" , dc, max_chunk); tr.speed_hz =mipi_dbi_spi_cmd_max_speed (spi, len);spi_message_init_with_transfers (&m, &tr,1 );
@@ -893,9 +891,7 @@ static int mipi_dbi_spi1_transfer(struct mipi_dbi *dbi, int dc, max_chunk = dbi->tx_buf9_len; dst16 = dbi->tx_buf9; -if (drm_debug_enabled (DRM_UT_DRIVER )) -pr_debug ("[drm:%s] dc=%d, max_chunk=%zu, transfers:\n" , - __func__, dc, max_chunk); +DRM_DEBUG_DRIVER ("dc=%d, max_chunk=%zu, transfers:\n" , dc, max_chunk); max_chunk =min (max_chunk /2 , len);
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v5 01/13] drm/mipi_dbi: Convert pr_debug calls to DRM_DEBUG_DRIVER
Hi Sean Paul.
Quoted 18 lines · Sean Paul
From: Sean Paul <[email protected]>
Use the drm logging helpers to output these messages to ensure they'll be included by the drm tracefs instance.
Signed-off-by Sean Paul <[email protected]>Changes in v5:
(quoted diff) +0
-2-Added to the set --- drivers/gpu/drm/drm_mipi_dbi.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-)drivers/gpu/drm/drm_mipi_dbi.c +0
-0
drivers/gpu/drm/drm_mipi_dbi.c static int mipi_dbi_spi1e_transfer(struct mipi_dbi *dbi, int dc, · line 772
Quoted 7 lines
drivers/gpu/drm/drm_mipi_dbi.c +1
-3@@ -769,9 +769,7 @@ static int mipi_dbi_spi1e_transfer(struct mipi_dbi *dbi, int dc,int i, ret;u8 *dst; -if (drm_debug_enabled (DRM_UT_DRIVER )) -pr_debug ("[drm:%s] dc=%d, max_chunk=%zu, transfers:\n" , - __func__, dc, max_chunk); +DRM_DEBUG_DRIVER ("dc=%d, max_chunk=%zu, transfers:\n" , dc, max_chunk);
Today we have drm_dbg(...) as the preferred logging method.
Sam
Quoted 21 lines
drivers/gpu/drm/drm_mipi_dbi.c +1
-3tr.speed_hz =mipi_dbi_spi_cmd_max_speed (spi, len);spi_message_init_with_transfers (&m, &tr,1 );@@ -893,9 +891,7 @@ static int mipi_dbi_spi1_transfer(struct mipi_dbi *dbi, int dc, max_chunk = dbi->tx_buf9_len; dst16 = dbi->tx_buf9; -if (drm_debug_enabled (DRM_UT_DRIVER )) -pr_debug ("[drm:%s] dc=%d, max_chunk=%zu, transfers:\n" , - __func__, dc, max_chunk); +DRM_DEBUG_DRIVER ("dc=%d, max_chunk=%zu, transfers:\n" , dc, max_chunk); max_chunk =min (max_chunk /2 , len);Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 2/13 drm/sil164: Convert dev_printk to drm_dev_dbg
In reply to Sean PaulFrom: Sean Paul <[email protected]>
Use the drm debug helper instead of dev_printk in order to leverage the upcoming tracefs support
Changes in v5:
-Added to the set
--- drivers/gpu/drm/i2c/sil164_drv.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-)
drivers/gpu/drm/i2c/sil164_drv.c +4 -8
@@ -43,11 +43,6 @@ struct sil164_priv {#define to_sil164_priv (x) \ ((struct sil164_priv *)to_encoder_slave(x)->slave_priv) -#define sil164_dbg (client, format, ...) do { \ - if (drm_debug_enabled(DRM_UT_KMS)) \ - dev_printk(KERN_DEBUG, &client->dev, \ - "%s: " format, __func__, ## __VA_ARGS__); \ - } while (0)#define sil164_info (client, format, ...) \ dev_info(&client->dev, format, __VA_ARGS__)#define sil164_err (client, format, ...) \
@@ -359,8 +354,8 @@ sil164_probe(struct i2c_client *client, const struct i2c_device_id *id)int rev =sil164_read (client,SIL164_REVISION );if (vendor !=0x1 || device !=0x6 ) { -sil164_dbg (client,"Unknown device %x:%x.%x\n" , - vendor, device, rev); +drm_dev_dbg (&client->dev,DRM_UT_KMS , +"Unknown device %x:%x.%x\n" , vendor, device, rev);return -ENODEV ; }
@@ -389,7 +384,8 @@ sil164_detect_slave(struct i2c_client *client) };if (i2c_transfer (adap, &msg,1 ) !=1 ) { -sil164_dbg (adap,"No dual-link slave found." ); +drm_dev_dbg (&adap->dev,DRM_UT_KMS , +"No dual-link slave found." );return NULL ; }
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 3/13 drm/i915/utils: Replace dev_printk with drm helpers
In reply to Sean PaulFrom: Sean Paul <[email protected]>
Use drm logging helpers to add support for the upcoming tracefs implementation.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/i915/i915_utils.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_utils.c +2 -3
@@ -30,10 +30,9 @@ __i915_printk(struct drm_i915_private *dev_priv, const char *level, vaf.va = &args;if (is_error) -dev_printk (level, kdev,"%pV" , &vaf); +drm_dev_printk (kdev, level,"%pV" , &vaf);else -dev_printk (level, kdev,"[" DRM_NAME ":%ps] %pV" , -__builtin_return_address (0 ), &vaf); +drm_err (&dev_priv->drm,"%pV" , &vaf);va_end (args);
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Sean Paul
reviewerPATCH v5 4/13 drm/msm/dpu: Replace definitions for dpu debug macros
In reply to Sean PaulFrom: Sean Paul <[email protected]>
The debug messages shouldn't be logged as errors when debug categories are enabled. Use the drm logging helpers to do the right thing
Changes in v5:
-Added to the set
--- drivers/gpu/drm/msm/disp/dpu1/dpu_kms.h | 20 ++++---------------- 1 file changed, 4 insertions(+), 16 deletions(-)
drivers/gpu/drm/msm/disp/dpu1/dpu_kms.h +4 -16
@@ -29,27 +29,15 @@* DPU_DEBUG - macro for kms/plane/crtc/encoder/connector logs * @fmt: Pointer to format string */ -#define DPU_DEBUG (fmt, ...) \ - do { \ - if (drm_debug_enabled(DRM_UT_KMS)) \ - DRM_DEBUG(fmt, ##__VA_ARGS__); \ - else \ - pr_debug(fmt, ##__VA_ARGS__); \ - } while (0) +#define DPU_DEBUG (fmt, ...) DRM_DEBUG_KMS(fmt, ##__VA_ARGS__)/** * DPU_DEBUG_DRIVER - macro for hardware driver logging * @fmt: Pointer to format string */ -#define DPU_DEBUG_DRIVER (fmt, ...) \ - do { \ - if (drm_debug_enabled(DRM_UT_DRIVER)) \ - DRM_ERROR(fmt, ##__VA_ARGS__); \ - else \ - pr_debug(fmt, ##__VA_ARGS__); \ - } while (0) - -#define DPU_ERROR (fmt, ...) pr_err("[dpu error]" fmt, ##__VA_ARGS__) +#define DPU_DEBUG_DRIVER (fmt, ...) DRM_DEBUG_DRIVER(fmt, ##__VA_ARGS__) + +#define DPU_ERROR (fmt, ...) DRM_ERROR(fmt, ##__VA_ARGS__)/** * ktime_compare_safe - compare two ktime structures
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 5/13 drm/print: rename drm_debug* to be more syslog-centric
In reply to Sean PaulFrom: Sean Paul <[email protected]>
In preparation for tracefs support, rename drm_debug related functions to reflect that it targets the syslog. This will allow us to selectively target syslog and/or tracefs.
No functional changes here.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/drm_print.c | 12 ++++++------ include/drm/drm_print.h | 13 +++++++++---- 2 files changed, 15 insertions(+), 10 deletions(-)
drivers/gpu/drm/drm_print.c +6 -6
@@ -37,11 +37,11 @@#include <drm/drm_print.h> /* -* __drm_debug: Enable debug output. +* __drm_debug_syslog: Enable debug output to system logs * Bitmask of DRM_UT_x. See include/drm/drm_print.h for details. */ -unsigned int __drm_debug; -EXPORT_SYMBOL (__drm_debug); +unsigned int __drm_debug_syslog; +EXPORT_SYMBOL (__drm_debug_syslog);MODULE_PARM_DESC (debug,"Enable debug output, where each bit enables a debug category.\n" "\t\tBit 0 (0x01) will enable CORE messages (drm core code)\n"
@@ -52,7 +52,7 @@ MODULE_PARM_DESC(debug, "Enable debug output, where each bit enables a debug cat"\t\tBit 5 (0x20) will enable VBL messages (vblank code)\n" "\t\tBit 7 (0x80) will enable LEASE messages (leasing code)\n" "\t\tBit 8 (0x100) will enable DP messages (displayport code)"); -module_param_named (debug, __drm_debug, int,0600 ); +module_param_named (debug, __drm_debug_syslog, int,0600 );void __drm_puts_coredump (struct drm_printer *p,const char *str) {
@@ -160,11 +160,11 @@ void __drm_printfn_info(struct drm_printer *p, struct va_format *vaf) }EXPORT_SYMBOL (__drm_printfn_info); -void __drm_printfn_debug (struct drm_printer *p,struct va_format *vaf) +void __drm_printfn_debug_syslog (struct drm_printer *p,struct va_format *vaf) {pr_debug ("%s %pV" , p->prefix, vaf); } -EXPORT_SYMBOL (__drm_printfn_debug); +EXPORT_SYMBOL (__drm_printfn_debug_syslog);void __drm_printfn_err (struct drm_printer *p,struct va_format *vaf) {
include/drm/drm_print.h +9 -4
@@ -35,7 +35,7 @@#include <drm/drm.h> /* Do *not* use outside of drm_print.[ch]! */ -extern unsigned int __drm_debug; +extern unsigned int __drm_debug_syslog;/** * DOC: print
@@ -85,7 +85,7 @@ void __drm_puts_coredump(struct drm_printer *p, const char *str);void __drm_printfn_seq_file (struct drm_printer *p,struct va_format *vaf);void __drm_puts_seq_file (struct drm_printer *p,const char *str);void __drm_printfn_info (struct drm_printer *p,struct va_format *vaf); -void __drm_printfn_debug (struct drm_printer *p,struct va_format *vaf); +void __drm_printfn_debug_syslog (struct drm_printer *p,struct va_format *vaf);void __drm_printfn_err (struct drm_printer *p,struct va_format *vaf);__printf (2 ,3 )
@@ -227,7 +227,7 @@ static inline struct drm_printer drm_info_printer(struct device *dev)static inline struct drm_printer drm_debug_printer (const char *prefix) {struct drm_printer p = { - .printfn = __drm_printfn_debug, + .printfn = __drm_printfn_debug_syslog, .prefix = prefix };return p;
@@ -319,9 +319,14 @@ enum drm_debug_category { DRM_UT_DRMRES = 0x200, }; +static inline bool drm_debug_syslog_enabled (enum drm_debug_category category) +{ +return unlikely (__drm_debug_syslog & category); +} +static inline bool drm_debug_enabled (enum drm_debug_category category) { -return unlikely (__drm_debug & category); +return drm_debug_syslog_enabled (category); }/*
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 6/13 drm/amd: Gate i2c transaction logs on drm_debug_syslog
In reply to Sean PaulFrom: Sean Paul <[email protected]>
Since the logs protected by these checks specifically target syslog, use the new drm_debug_syslog_enabled() call to avoid triggering these prints when only trace is enabled.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c +2 -2
@@ -233,7 +233,7 @@ static uint32_t smu_v11_0_i2c_transmit(struct i2c_adapter *control,DRM_DEBUG_DRIVER ("I2C_Transmit(), address = %x, bytes = %d , data: " , (uint16_t )address, numbytes); -if (drm_debug_enabled (DRM_UT_DRIVER )) { +if (drm_debug_syslog_enabled (DRM_UT_DRIVER )) {print_hex_dump (KERN_INFO ,"data: " ,DUMP_PREFIX_NONE ,16 ,1 , data, numbytes, false); }
@@ -387,7 +387,7 @@ static uint32_t smu_v11_0_i2c_receive(struct i2c_adapter *control,DRM_DEBUG_DRIVER ("I2C_Receive(), address = %x, bytes = %d, data :" , (uint16_t )address, bytes_received); -if (drm_debug_enabled (DRM_UT_DRIVER )) { +if (drm_debug_syslog_enabled (DRM_UT_DRIVER )) {print_hex_dump (KERN_INFO ,"data: " ,DUMP_PREFIX_NONE ,16 ,1 , data, bytes_received, false); }
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Re: [PATCH v5 06/13] drm/amd: Gate i2c transaction logs on drm_debug_syslog
Am 08.06.20 um 23:04 schrieb Sean Paul:
Quoted 5 lines
From: Sean Paul <[email protected]>
Since the logs protected by these checks specifically target syslog, use the new
drm_debug_syslog_enabled()call to avoid triggering these prints when only trace is enabled.
Mhm, of hand that doesn't looks like something which belongs into the syslog in the first place. Maybe convert it into a trace point instead?
Quoted 2 lines
Signed-off-by Sean Paul <[email protected]>
Quoted 21 lines
Changes in v5:
(quoted diff) +0
-2-Added to the set --- drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c +1
-1@@ -233,7 +233,7 @@ static uint32_t smu_v11_0_i2c_transmit(struct i2c_adapter *control,DRM_DEBUG_DRIVER ("I2C_Transmit(), address = %x, bytes = %d , data: " , (uint16_t )address, numbytes); -if (drm_debug_enabled (DRM_UT_DRIVER )) { +if (drm_debug_syslog_enabled (DRM_UT_DRIVER )) {print_hex_dump (KERN_INFO ,"data: " ,DUMP_PREFIX_NONE ,16 ,1 , data, numbytes, false); }
drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c static uint32_t smu_v11_0_i2c_receive(struct i2c_adapter *control, · line 393
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c +1
-1@@ -387,7 +387,7 @@ static uint32_t smu_v11_0_i2c_receive(struct i2c_adapter *control,DRM_DEBUG_DRIVER ("I2C_Receive(), address = %x, bytes = %d, data :" , (uint16_t )address, bytes_received); -if (drm_debug_enabled (DRM_UT_DRIVER )) { +if (drm_debug_syslog_enabled (DRM_UT_DRIVER )) {print_hex_dump (KERN_INFO ,"data: " ,DUMP_PREFIX_NONE ,16 ,1 , data, bytes_received, false); }
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Sean Paul
reviewerPATCH v5 7/13 drm/etnaviv: Change buffer dump checks to target syslog
In reply to Sean PaulFrom: Sean Paul <[email protected]>
Since the logs protected by these checks specifically target syslog, use the new drm_debug_syslog_enabled() call to avoid triggering these prints when only trace is enabled.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/etnaviv/etnaviv_buffer.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/etnaviv/etnaviv_buffer.c +4 -4
@@ -353,7 +353,7 @@ void etnaviv_buffer_queue(struct etnaviv_gpu *gpu, u32 exec_state,lockdep_assert_held (&gpu->lock); -if (drm_debug_enabled (DRM_UT_DRIVER )) +if (drm_debug_syslog_enabled (DRM_UT_DRIVER ))etnaviv_buffer_dump (gpu, buffer,0 ,0x50 ); link_target = etnaviv_cmdbuf_get_va(cmdbuf,
@@ -509,13 +509,13 @@ void etnaviv_buffer_queue(struct etnaviv_gpu *gpu, u32 exec_state,etnaviv_cmdbuf_get_va (buffer, &gpu->mmu_context->cmdbuf_mapping) + buffer->user_size - 4); -if (drm_debug_enabled (DRM_UT_DRIVER )) +if (drm_debug_syslog_enabled (DRM_UT_DRIVER ))pr_info ("stream link to 0x%08x @ 0x%08x %p\n" , return_target,etnaviv_cmdbuf_get_va (cmdbuf, &gpu->mmu_context->cmdbuf_mapping), cmdbuf->vaddr); -if (drm_debug_enabled (DRM_UT_DRIVER )) { +if (drm_debug_syslog_enabled (DRM_UT_DRIVER )) {print_hex_dump (KERN_INFO ,"cmd " ,DUMP_PREFIX_OFFSET ,16 ,4 , cmdbuf->vaddr, cmdbuf->size,0 );
@@ -534,6 +534,6 @@ void etnaviv_buffer_queue(struct etnaviv_gpu *gpu, u32 exec_state,VIV_FE_LINK_HEADER_PREFETCH (link_dwords), link_target); -if (drm_debug_enabled (DRM_UT_DRIVER )) +if (drm_debug_syslog_enabled (DRM_UT_DRIVER ))etnaviv_buffer_dump (gpu, buffer,0 ,0x50 ); }
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 8/13 drm/nouveau: Change debug checks to specifically target syslog
In reply to Sean PaulFrom: Sean Paul <[email protected]>
Since the logs protected by these checks specifically target syslog, use the new drm_debug_syslog_enabled() call to avoid triggering these prints when only trace is enabled.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/nouveau/dispnv50/disp.h | 4 ++-- drivers/gpu/drm/nouveau/nouveau_drv.h | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/nouveau/dispnv50/disp.h +2 -2
@@ -85,14 +85,14 @@ extern const u64 wndwc57e_modifiers[];#define evo_mthd (p, m, s) do { \ const u32 _m = (m), _s = (s); \ - if (drm_debug_enabled(DRM_UT_KMS)) \ + if (drm_debug_syslog_enabled(DRM_UT_KMS)) \ pr_err("%04x %d %s\n", _m, _s, __func__); \ *((p)++) = ((_s << 18) | _m); \ } while(0)#define evo_data (p, d) do { \ const u32 _d = (d); \ - if (drm_debug_enabled(DRM_UT_KMS)) \ + if (drm_debug_syslog_enabled(DRM_UT_KMS)) \ pr_err("\t%08x\n", _d); \ *((p)++) = _d; \ } while(0)
drivers/gpu/drm/nouveau/nouveau_drv.h +2 -2
@@ -257,11 +257,11 @@ void nouveau_drm_device_remove(struct drm_device *dev); #define NV_INFO(drm,f,a...) NV_PRINTK(info, &(drm)->client, f, ##a) #define NV_DEBUG(drm,f,a...) do { \ - if (drm_debug_enabled(DRM_UT_DRIVER)) \ + if (drm_debug_syslog_enabled(DRM_UT_DRIVER)) \ NV_PRINTK(info, &(drm)->client, f, ##a); \ } while(0) #define NV_ATOMIC(drm,f,a...) do { \ - if (drm_debug_enabled(DRM_UT_ATOMIC)) \ + if (drm_debug_syslog_enabled(DRM_UT_ATOMIC)) \ NV_PRINTK(info, &(drm)->client, f, ##a); \ } while(0)
Signature · 1 line
Sean Paul, Software Engineer, Google / Chromium OS
Sean Paul
reviewerPATCH v5 9/13 drm/i915: Change infoframe debug checks to specify syslog
In reply to Sean PaulFrom: Sean Paul <[email protected]>
Since the logs protected by these checks specifically target syslog, use the new drm_debug_syslog_enabled() call to avoid triggering these prints when only trace is enabled.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/i915/display/intel_display.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/display/intel_display.c +2 -2
@@ -12876,7 +12876,7 @@ static void intel_dump_infoframe(struct drm_i915_private *dev_priv, const union hdmi_infoframe *frame) { -if (!drm_debug_enabled (DRM_UT_KMS )) +if (!drm_debug_syslog_enabled (DRM_UT_KMS ))return ;hdmi_infoframe_log (KERN_DEBUG , dev_priv->drm.dev, frame);
@@ -13519,7 +13519,7 @@ pipe_config_infoframe_mismatch(struct drm_i915_private *dev_priv, const union hdmi_infoframe *b) {if (fastset) { -if (!drm_debug_enabled (DRM_UT_KMS )) +if (!drm_debug_syslog_enabled (DRM_UT_KMS ))return ; drm_dbg_kms(&dev_priv->drm,
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Sean Paul
reviewerPATCH v5 10/13 drm/print: Add drm_debug_category_printer
In reply to Sean PaulFrom: Sean Paul <[email protected]>
This patch adds a new printer which will select the appropriate output for a given debug category. Currently there is only one output target, which is syslog. However in the future we'll have tracefs and it will be useful to print to syslog, tracefs, or both. Drivers just need to create the printer for the appropriate category and the printer will decide where to send the output.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/drm_print.c | 5 +++++ include/drm/drm_print.h | 28 ++++++++++++++++++++++++++++ 2 files changed, 33 insertions(+)
drivers/gpu/drm/drm_print.c +5 -0
@@ -172,6 +172,11 @@ void __drm_printfn_err(struct drm_printer *p, struct va_format *vaf) }EXPORT_SYMBOL (__drm_printfn_err); +void __drm_printfn_noop (struct drm_printer *p,struct va_format *vaf) +{ +} +EXPORT_SYMBOL (__drm_printfn_noop); +/** * drm_puts - print a const string to a &drm_printer stream * @p: the &drm printer
include/drm/drm_print.h +28 -0
@@ -87,6 +87,7 @@ void __drm_puts_seq_file(struct drm_printer *p, const char *str);void __drm_printfn_info (struct drm_printer *p,struct va_format *vaf);void __drm_printfn_debug_syslog (struct drm_printer *p,struct va_format *vaf);void __drm_printfn_err (struct drm_printer *p,struct va_format *vaf); +void __drm_printfn_noop (struct drm_printer *p,struct va_format *vaf); __printf(2, 3)void drm_printf (struct drm_printer *p,const char *f, ...);
@@ -329,6 +330,33 @@ static inline bool drm_debug_enabled(enum drm_debug_category category)return drm_debug_syslog_enabled (category); } +/** +* drm_debug_category_printer - construct a &drm_printer that outputs to +* pr_debug() if enabled for the given category. +* @category: the DRM_UT_* message category this message belongs to +* @prefix: trace output prefix +* +* RETURNS: +* The &drm_printer object +*/ +static inline struct drm_printer +drm_debug_category_printer (enum drm_debug_category category, +const char *prefix) +{ +struct drm_printer p = { + .prefix = prefix + }; + +if (drm_debug_syslog_enabled (category)) { + p.printfn = __drm_printfn_debug_syslog; + }else { +WARN (1 ,"Debug category %d is inactive." , category); + p.printfn = __drm_printfn_noop; + } + +return p; +} +/* * struct device based logging *
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 11/13 drm/mst: Convert debug printers to debug category printers
In reply to Sean PaulFrom: Sean Paul <[email protected]>
The printers in dp_mst are meant to be gated on DRM_UT_DP, so use the debug category printer to avoid dumping mst transactions to the wrong place.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/drm_dp_mst_topology.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/gpu/drm/drm_dp_mst_topology.c +6 -3
@@ -1203,7 +1203,8 @@ static int drm_dp_mst_wait_tx_reply(struct drm_dp_mst_branch *mstb, } out:if (unlikely (ret == -EIO ) &&drm_debug_enabled (DRM_UT_DP )) { -struct drm_printer p =drm_debug_printer (DBG_PREFIX ); +struct drm_printer p =drm_debug_category_printer (DRM_UT_DP , +DBG_PREFIX );drm_dp_mst_dump_sideband_msg_tx (&p, txmsg); }
@@ -2739,7 +2740,8 @@ static int process_single_tx_qlock(struct drm_dp_mst_topology_mgr *mgr, ret =drm_dp_send_sideband_msg (mgr, up, chunk, idx);if (unlikely (ret) &&drm_debug_enabled (DRM_UT_DP )) { -struct drm_printer p =drm_debug_printer (DBG_PREFIX ); +struct drm_printer p =drm_debug_category_printer (DRM_UT_DP , +DBG_PREFIX );drm_printf (&p,"sideband msg failed to send\n" );drm_dp_mst_dump_sideband_msg_tx (&p, txmsg);
@@ -2783,7 +2785,8 @@ static void drm_dp_queue_down_tx(struct drm_dp_mst_topology_mgr *mgr,list_add_tail (&txmsg->next, &mgr->tx_msg_downq);if (drm_debug_enabled (DRM_UT_DP )) { -struct drm_printer p =drm_debug_printer (DBG_PREFIX ); +struct drm_printer p =drm_debug_category_printer (DRM_UT_DP , +DBG_PREFIX );drm_dp_mst_dump_sideband_msg_tx (&p, txmsg); }
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerPATCH v5 12/13 drm/i915: Use debug category printer for welcome message
In reply to Sean PaulFrom: Sean Paul <[email protected]>
The welcome printer is meant to be gated on DRM_UT_DRIVER, so use the debug category printer to avoid dumping the message in the wrong place.
Changes in v5:
-Added to the set
--- drivers/gpu/drm/i915/i915_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_drv.c +2 -1
@@ -876,7 +876,8 @@ static void i915_driver_unregister(struct drm_i915_private *dev_priv)static void i915_welcome_messages (struct drm_i915_private *dev_priv) {if (drm_debug_enabled (DRM_UT_DRIVER )) { -struct drm_printer p =drm_debug_printer ("i915 device info:" ); +struct drm_printer p =drm_debug_category_printer (DRM_UT_DRIVER , +"i915 device info:" ); drm_printf(&p, "pciid=0x%04x rev=0x%02x platform=%s (subplatform=0x%x) gen=%i\n", INTEL_DEVID(dev_priv),
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Sean Paul
reviewerPATCH v5 13/13 drm/print: Add tracefs support to the drm logging helpers
In reply to Sean PaulFrom: Sean Paul <[email protected]>
This patch adds a new module parameter called drm.trace which accepts the same mask as drm.debug. When a debug category is enabled, log messages will be put in a new tracefs instance called drm for consumption.
Using the new tracefs instance will allow distros to enable drm logging in production without impacting performance or spamming the system logs.
Changes in v5:
-Re-write to use trace_array and the tracefs instance support
--- Documentation/gpu/drm-uapi.rst | 6 + drivers/gpu/drm/drm_drv.c | 3 + drivers/gpu/drm/drm_print.c | 209 ++++++++++++++++++++++++++++----- include/drm/drm_print.h | 63 ++++++++-- 4 files changed, 241 insertions(+), 40 deletions(-)
Documentation/gpu/drm-uapi.rst +6 -0
@@ -312,6 +312,12 @@ Debugfs Support .. kernel-doc:: drivers/gpu/drm/drm_debugfs.c :export: +DRM Tracing +--------------- + +.. kernel-doc:: drivers/gpu/drm/drm_print.c + :doc: DRM Tracing + Sysfs Support =============
drivers/gpu/drm/drm_drv.c +3 -0
@@ -1108,12 +1108,15 @@ static void drm_core_exit(void)drm_sysfs_destroy ();idr_destroy (&drm_minors_idr);drm_connector_ida_destroy (); +drm_trace_cleanup (); }static int __init drm_core_init (void ) {int ret; +drm_trace_init (); +drm_connector_ida_init ();idr_init (&drm_minors_idr);
drivers/gpu/drm/drm_print.c +180 -29
257 lines not on this page · raw message
include/drm/drm_print.h +52 -11
136 lines not on this page · raw message
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
kernel test robot
build botRe: [PATCH v5 13/13] drm/print: Add tracefs support to the drm logging helpers
I love your patch! Perhaps something to improve:
Hi Sean,
I love your patch! Perhaps something to improve:
[auto build test WARNING on drm-intel/for-linux-next]
[also build test WARNING on drm-exynos/exynos-drm-next tegra-drm/drm/tegra/for-next drm-tip/drm-tip linus/master v5.7] [cannot apply to drm/drm-next]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]
url: https://github.com/0day-ci/linux/commits/Sean-Paul/drm-trace-Mirror-DRM-debug-logs-to-tracefs/20200609-050750 base: git://anongit.freedesktop.org/drm-intel for-linux-next config: riscv-allyesconfig (attached as .config) compiler: riscv64-linux-gcc (GCC) 9.3.0
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross ARCH=riscv
If you fix the issue, kindly add following tag as appropriate
All warnings (new ones prefixed by >>, old ones prefixed by <<):
drivers/gpu/drm/drm_print.c: In function 'drm_trace_init':
>> drivers/gpu/drm/drm_print.c:459:6: warning: old-style function definition [-Wold-style-definition]
459 | void drm_trace_init()
| ^~~~~~~~~~~~~~drivers/gpu/drm/drm_print.c: In function 'drm_trace_cleanup':drivers/gpu/drm/drm_print.c:492:6: warning: old-style function definition [-Wold-style-definition] 492 | void drm_trace_cleanup()
| ^~~~~~~~~~~~~~~~~
vim +459 drivers/gpu/drm/drm_print.c
452
453 /**
454 * drm_trace_init - initializes the drm trace array
455 *
456 * This function fetches (or creates) the drm trace array. This should be called
457 * once on drm subsystem creation and matched with drm_trace_cleanup().
458 */
> 459 void drm_trace_init()
460 {
461 trace_arr = trace_array_get_by_name("drm");
462 if (!trace_arr)
463 return;
464 }
465 EXPORT_SYMBOL(drm_trace_init);
466
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
kernel test robot
build botRe: [PATCH v5 13/13] drm/print: Add tracefs support to the drm logging helpers
In reply to Sean PaulI love your patch! Perhaps something to improve:
Hi Sean,
I love your patch! Perhaps something to improve:
[auto build test WARNING on drm-intel/for-linux-next]
[also build test WARNING on drm-exynos/exynos-drm-next tegra-drm/drm/tegra/for-next drm-tip/drm-tip linus/master v5.7 next-20200608] [cannot apply to drm/drm-next]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]
url: https://github.com/0day-ci/linux/commits/Sean-Paul/drm-trace-Mirror-DRM-debug-logs-to-tracefs/20200609-050750 base: git://anongit.freedesktop.org/drm-intel for-linux-next config: x86_64-randconfig-s021-20200607 (attached as .config) compiler: gcc-9 (Debian 9.3.0-13) 9.3.0 reproduce: # apt-get install sparse # sparse version: v0.6.1-247-gcadbd124-dirty # save the attached .config to linux build tree make W=1 C=1 ARCH=x86_64 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__'
If you fix the issue, kindly add following tag as appropriate
sparse warnings: (new ones prefixed by >>)
>> drivers/gpu/drm/drm_print.c:459:21: sparse: sparse: non-ANSI function declaration of function 'drm_trace_init' >> drivers/gpu/drm/drm_print.c:492:24: sparse: sparse: non-ANSI function declaration of function 'drm_trace_cleanup'
Please review and possibly fold the followup patch.
vim +/drm_trace_init +459 drivers/gpu/drm/drm_print.c
452
453 /**
454 * drm_trace_init - initializes the drm trace array
455 *
456 * This function fetches (or creates) the drm trace array. This should be called
457 * once on drm subsystem creation and matched with drm_trace_cleanup().
458 */
> 459 void drm_trace_init()
460 {
461 trace_arr = trace_array_get_by_name("drm");
462 if (!trace_arr)
463 return;
464 }
465 EXPORT_SYMBOL(drm_trace_init);
466
467 /**
468 * drm_trace_printf - adds an entry to the drm tracefs instance
469 * @format: printf format of the message to add to the trace
470 *
471 * This function adds a new entry in the drm tracefs instance
472 */
473 void drm_trace_printf(const char *format, ...)
474 {
475 struct va_format vaf;
476 va_list args;
477
478 va_start(args, format);
479 vaf.fmt = format;
480 vaf.va = &args;
481 trace_array_printk(trace_arr, _THIS_IP_, "%pV", &vaf);
482 va_end(args);
483 }
484
485 /**
486 * drm_trace_cleanup - destroys the drm trace array
487 *
488 * This function destroys the drm trace array created with drm_trace_init. This
489 * should be called once on drm subsystem close and matched with
490 * drm_trace_init().
491 */
> 492 void drm_trace_cleanup()---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
Sean Paul
reviewerPATCH v6 13/13 drm/print: Add tracefs support to the drm logging helpers
From: Sean Paul <[email protected]>
This patch adds a new module parameter called drm.trace which accepts the same mask as drm.debug. When a debug category is enabled, log messages will be put in a new tracefs instance called drm for consumption.
Using the new tracefs instance will allow distros to enable drm logging in production without impacting performance or spamming the system logs.
Changes in v5:
-Re-write to use trace_array and the tracefs instance support
Changes in v6:
-Fix old-style function definition warnings
--- Documentation/gpu/drm-uapi.rst | 6 + drivers/gpu/drm/drm_drv.c | 3 + drivers/gpu/drm/drm_print.c | 209 ++++++++++++++++++++++++++++----- include/drm/drm_print.h | 63 ++++++++-- 4 files changed, 241 insertions(+), 40 deletions(-)
Documentation/gpu/drm-uapi.rst +6 -0
@@ -312,6 +312,12 @@ Debugfs Support .. kernel-doc:: drivers/gpu/drm/drm_debugfs.c :export: +DRM Tracing +--------------- + +.. kernel-doc:: drivers/gpu/drm/drm_print.c + :doc: DRM Tracing + Sysfs Support =============
drivers/gpu/drm/drm_drv.c +3 -0
@@ -1108,12 +1108,15 @@ static void drm_core_exit(void)drm_sysfs_destroy ();idr_destroy (&drm_minors_idr);drm_connector_ida_destroy (); +drm_trace_cleanup (); }static int __init drm_core_init (void ) {int ret; +drm_trace_init (); +drm_connector_ida_init ();idr_init (&drm_minors_idr);
drivers/gpu/drm/drm_print.c +180 -29
256 lines not on this page · raw message
include/drm/drm_print.h +52 -11
136 lines not on this page · raw message
Signature · 6 lines
Sean Paul, Software Engineer, Google / Chromium OS _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v6 13/13] drm/print: Add tracefs support to the drm logging helpers
Quoted 7 lines · On Tue, 9 Jun 2020 11:49:19 -0400 Sean Paul <[email protected]> wrote:
drivers/gpu/drm/drm_print.c +7
-0+/** +* drm_trace_printf - adds an entry to the drm tracefs instance +* @format: printf format of the message to add to the trace +* +* This function adds a new entry in the drm tracefs instance +*/ +void drm_trace_printf (const char *format, ...)
drivers/gpu/drm/drm_print.c void drm_print_regset32(struct drm_printer *p, struct debugfs_regset32 *regset) · line 483
Quoted 10 lines
drivers/gpu/drm/drm_print.c +10
-0+{ +struct va_format vaf; +va_list args; + +va_start (args, format); + vaf.fmt = format; + vaf.va = &args; +trace_array_printk (trace_arr, _THIS_IP_,"%pV" , &vaf); +va_end (args); +}
The only issue I have with this patch is the above. I really dislike the use of trace_array_print(), as that is made to be for debugging and not something that should be in a production kernel.
Ideally, every instance should just pass the data it wants to record, and you add it to a trace event. There's already a drm trace subsystem, how would this be different. Perhaps create a drm_log subsystem, and you only need to have your instance enable it?
I guess I'm still confused to why this is needed over just having trace events? What's special about this case?
-- Steve
Sean Paul
reviewerRe: [PATCH v6 13/13] drm/print: Add tracefs support to the drm logging helpers
Quoted 26 lines · Steven Rostedt
Quoted 7 lines · On Tue, 9 Jun 2020 11:49:19 -0400 Sean Paul <[email protected]> wrote:
drivers/gpu/drm/drm_print.c +7
-0+/** +* drm_trace_printf - adds an entry to the drm tracefs instance +* @format: printf format of the message to add to the trace +* +* This function adds a new entry in the drm tracefs instance +*/ +void drm_trace_printf (const char *format, ...)drivers/gpu/drm/drm_print.c void drm_print_regset32(struct drm_printer *p, struct debugfs_regset32 *regset) · line 483
Quoted 10 lines
drivers/gpu/drm/drm_print.c +10
-0+{ +struct va_format vaf; +va_list args; + +va_start (args, format); + vaf.fmt = format; + vaf.va = &args; +trace_array_printk (trace_arr, _THIS_IP_,"%pV" , &vaf); +va_end (args); +}The only issue I have with this patch is the above. I really dislike the use of
trace_array_print(), as that is made to be for debugging and not something that should be in a production kernel.
Quoted 8 lines
Ideally, every instance should just pass the data it wants to record, and you add it to a trace event. There's already a drm trace subsystem, how would this be different. Perhaps create a
drm_logsubsystem, and you only need to have your instance enable it?I guess I'm still confused to why this is needed over just having trace events? What's special about this case?
Hi Steve,
In v2 I had added trace events throughout drm in this manner. See
https://lists.freedesktop.org/archives/dri-devel/2019-November/243318.html
The feedback I received was that exposing the events in such a structured way would bind us to the trace formats and make things more difficult moving forward.
Aside from the UAPI concerns, wearing my ChromeOS hat (with full knowledge that most upstream folks won't care about this), this solution also misses on making problems easier to diagnose from logs/user feedback. It requires that driver developers add tracing beside their logging, and makes backporting upstream code even harder than it is right now. If we can intercept the existing and new logging at the source, it's much easier to ensure the flight recorder is accurate. We've been shipping this (v4) in CrOS for ~months now and it's been invaluable. With this experience in mind, I'd be really hesitant to change course.
Sean
Quoted 2 lines
-- Steve
kernel test robot
build bot[RFC PATCH] drm/print: trace_arr can be static
In reply to Sean PaulSigned-off-by: kernel test robot <[email protected]>
--- drm_print.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/drm_print.c +1 -1
@@ -69,7 +69,7 @@ MODULE_PARM_DESC(trace, DEBUG_PARM_DESC("tracefs"));module_param_named (trace, __drm_debug_trace, int,0600 );#ifdef CONFIG_TRACING -struct trace_array *trace_arr; +static struct trace_array *trace_arr;#endif void __drm_puts_coredump (struct drm_printer *p,const char *str)
Re: [PATCH v5 13/13] drm/print: Add tracefs support to the drm logging helpers
In reply to Sean PaulQuoted 95 lines · On Mon, 8 Jun 2020 17:05:03 -0400 Sean Paul <[email protected]> wrote:
From: Sean Paul <[email protected]>
This patch adds a new module parameter called drm.trace which accepts the same mask as drm.debug. When a debug category is enabled, log messages will be put in a new tracefs instance called drm for consumption.
Using the new tracefs instance will allow distros to enable drm logging in production without impacting performance or spamming the system logs.
Cc Daniel Vetter <[email protected]>Cc David Airlie <[email protected]>Cc Jani Nikula <[email protected]>Cc Joonas Lahtinen <[email protected]>Cc Pekka Paalanen <[email protected]>Cc Rob Clark <[email protected]>Cc Steven Rostedt <[email protected]>Cc Thomas Zimmermann <[email protected]>Cc Ville Syrjälä <[email protected]>Cc Chris Wilson <[email protected]>Cc Steven Rostedt <[email protected]>Signed-off-by Sean Paul <[email protected]>Changes in v5:
(quoted diff) +0
-2-Re-write to use trace_array and the tracefs instance support --- Documentation/gpu/drm-uapi.rst | 6 + drivers/gpu/drm/drm_drv.c | 3 + drivers/gpu/drm/drm_print.c | 209 ++++++++++++++++++++++++++++----- include/drm/drm_print.h | 63 ++++++++-- 4 files changed, 241 insertions(+), 40 deletions(-)Documentation/gpu/drm-uapi.rst +6
-0@@ -312,6 +312,12 @@ Debugfs Support .. kernel-doc:: drivers/gpu/drm/drm_debugfs.c :export: +DRM Tracing +--------------- + +.. kernel-doc:: drivers/gpu/drm/drm_print.c + :doc: DRM Tracing + Sysfs Support =============drivers/gpu/drm/drm_drv.c +3
-0@@ -1108,12 +1108,15 @@ static void drm_core_exit(void)drm_sysfs_destroy ();idr_destroy (&drm_minors_idr);drm_connector_ida_destroy (); +drm_trace_cleanup (); }static int __init drm_core_init (void ) {int ret; +drm_trace_init (); +drm_connector_ida_init ();idr_init (&drm_minors_idr);drivers/gpu/drm/drm_print.c +1
-5@@ -31,6 +31,7 @@#include <linux/moduleparam.h> #include <linux/seq_file.h> #include <linux/slab.h> +#include <linux/trace.h> #include <drm/drm.h> #include <drm/drm_drv.h> @@ -43,17 +44,34 @@unsigned int __drm_debug_syslog;EXPORT_SYMBOL (__drm_debug_syslog); -MODULE_PARM_DESC (debug,"Enable debug output, where each bit enables a debug category.\n" -"\t\tBit 0 (0x01) will enable CORE messages (drm core code)\n" -"\t\tBit 1 (0x02) will enable DRIVER messages (drm controller code)\n" -"\t\tBit 2 (0x04) will enable KMS messages (modesetting code)\n" -"\t\tBit 3 (0x08) will enable PRIME messages (prime code)\n"
drivers/gpu/drm/drm_print.c line 52
Quoted 10 lines
drivers/gpu/drm/drm_print.c +6
-4-"\t\tBit 4 (0x10) will enable ATOMIC messages (atomic code)\n" -"\t\tBit 5 (0x20) will enable VBL messages (vblank code)\n" -"\t\tBit 7 (0x80) will enable LEASE messages (leasing code)\n" -"\t\tBit 8 (0x100) will enable DP messages (displayport code)"); +/* +* __drm_debug_trace: Enable debug output in drm tracing instance. +* Bitmask of DRM_UT_x. See include/drm/drm_print.h for details. +*/ +unsigned int __drm_debug_trace; +EXPORT_SYMBOL (__drm_debug_trace);
Hi!
Might distributions perhaps want to set a default value for this via Kconfig? Or could setting it via sysfs happen early enough to diagnose e.g. Plymouth problems?
Or maybe there is nothing to see from early boot?
The general usefulness of this feature depends on whether people actually run with it enabled.
Quoted 17 lines
drivers/gpu/drm/drm_print.c +15
-0+ +#define DEBUG_PARM_DESC (dst) \ +"Enable debug output to " dst ", where each bit enables a debug category.\n" \ +"\t\tBit 0 (0x01) will enable CORE messages (drm core code)\n" \ +"\t\tBit 1 (0x02) will enable DRIVER messages (drm controller code)\n" \ +"\t\tBit 2 (0x04) will enable KMS messages (modesetting code)\n" \ +"\t\tBit 3 (0x08) will enable PRIME messages (prime code)\n" \ +"\t\tBit 4 (0x10) will enable ATOMIC messages (atomic code)\n" \ +"\t\tBit 5 (0x20) will enable VBL messages (vblank code)\n" \ +"\t\tBit 7 (0x80) will enable LEASE messages (leasing code)\n" \ +"\t\tBit 8 (0x100) will enable DP messages (displayport code)" + +MODULE_PARM_DESC (debug,DEBUG_PARM_DESC ("syslog" ));module_param_named (debug, __drm_debug_syslog, int,0600 ); +MODULE_PARM_DESC (trace,DEBUG_PARM_DESC ("tracefs" )); +module_param_named (trace, __drm_debug_trace, int,0600 );
...
Quoted 31 lines
drivers/gpu/drm/drm_print.c +31
-0+ +/** +* DOC: DRM Tracing +* +* *tl;dr* DRM tracing is a lightweight alternative to traditional DRM debug +* logging. +* +* While DRM logging is quite convenient when reproducing a specific issue, it +* doesn't help when something goes wrong unexpectedly. There are a couple +* reasons why one does not want to enable DRM logging at all times: +* +* 1. We don't want to overwhelm syslog with drm spam, others have to use it too +* 2. Console logging is slow +* +* DRM tracing aims to solve both these problems. +* +* To use DRM tracing, set the drm.trace module parameter (via cmdline or sysfs) +* to a DRM debug category mask (this is a bitmask of &drm_debug_category +* values): +* :: +* +* eg: echo 0x106 > /sys/module/drm/parameters/trace +* +* Once active, all log messages in the specified categories will be written to +* the DRM trace. Once at capacity, the trace will overwrite old messages with +* new ones. At any point, one can read the trace file to extract the previous N +* DRM messages: +* :: +* +* eg: cat /sys/kernel/tracing/instances/drm/trace +*
drivers/gpu/drm/drm_print.c void drm_print_regset32(struct drm_printer *p, struct debugfs_regset32 *regset) · line 448
Quoted 10 lines
drivers/gpu/drm/drm_print.c +10
-0+* Considerations +* ************** +* The trace is subsystem wide, so if you have multiple devices active, they +* will be adding logs to the same trace. +* +* The contents of the DRM Trace are **not** considered UABI. **DO NOT depend on +* the values of these traces in your userspace.** These traces are intended for +* entertainment purposes only. The contents of these logs carry no warranty, +* expressed or implied. +*/
Sounds good to me!
This part is:
Thanks,
pq
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Sean Paul
reviewerRe: [PATCH v5 13/13] drm/print: Add tracefs support to the drm logging helpers
Quoted 103 lines · Pekka Paalanen
Quoted 99 lines · On Mon, 8 Jun 2020 17:05:03 -0400 Sean Paul <[email protected]> wrote:
From: Sean Paul <[email protected]>
This patch adds a new module parameter called drm.trace which accepts the same mask as drm.debug. When a debug category is enabled, log messages will be put in a new tracefs instance called drm for consumption.
Using the new tracefs instance will allow distros to enable drm logging in production without impacting performance or spamming the system logs.
Cc Daniel Vetter <[email protected]>Cc David Airlie <[email protected]>Cc Jani Nikula <[email protected]>Cc Joonas Lahtinen <[email protected]>Cc Pekka Paalanen <[email protected]>Cc Rob Clark <[email protected]>Cc Steven Rostedt <[email protected]>Cc Thomas Zimmermann <[email protected]>Cc Ville Syrjälä <[email protected]>Cc Chris Wilson <[email protected]>Cc Steven Rostedt <[email protected]>Signed-off-by Sean Paul <[email protected]>Changes in v5:
(quoted diff) +0
-2-Re-write to use trace_array and the tracefs instance support --- Documentation/gpu/drm-uapi.rst | 6 + drivers/gpu/drm/drm_drv.c | 3 + drivers/gpu/drm/drm_print.c | 209 ++++++++++++++++++++++++++++----- include/drm/drm_print.h | 63 ++++++++-- 4 files changed, 241 insertions(+), 40 deletions(-)Documentation/gpu/drm-uapi.rst +6
-0@@ -312,6 +312,12 @@ Debugfs Support .. kernel-doc:: drivers/gpu/drm/drm_debugfs.c :export: +DRM Tracing +--------------- + +.. kernel-doc:: drivers/gpu/drm/drm_print.c + :doc: DRM Tracing + Sysfs Support =============drivers/gpu/drm/drm_drv.c +3
-0@@ -1108,12 +1108,15 @@ static void drm_core_exit(void)drm_sysfs_destroy ();idr_destroy (&drm_minors_idr);drm_connector_ida_destroy (); +drm_trace_cleanup (); }static int __init drm_core_init (void ) {int ret; +drm_trace_init (); +drm_connector_ida_init ();idr_init (&drm_minors_idr);drivers/gpu/drm/drm_print.c +1
-9@@ -31,6 +31,7 @@#include <linux/moduleparam.h> #include <linux/seq_file.h> #include <linux/slab.h> +#include <linux/trace.h> #include <drm/drm.h> #include <drm/drm_drv.h> @@ -43,17 +44,34 @@unsigned int __drm_debug_syslog;EXPORT_SYMBOL (__drm_debug_syslog); -MODULE_PARM_DESC (debug,"Enable debug output, where each bit enables a debug category.\n" -"\t\tBit 0 (0x01) will enable CORE messages (drm core code)\n" -"\t\tBit 1 (0x02) will enable DRIVER messages (drm controller code)\n" -"\t\tBit 2 (0x04) will enable KMS messages (modesetting code)\n" -"\t\tBit 3 (0x08) will enable PRIME messages (prime code)\n" -"\t\tBit 4 (0x10) will enable ATOMIC messages (atomic code)\n" -"\t\tBit 5 (0x20) will enable VBL messages (vblank code)\n" -"\t\tBit 7 (0x80) will enable LEASE messages (leasing code)\n" -"\t\tBit 8 (0x100) will enable DP messages (displayport code)" );
Quoted 8 lines
drivers/gpu/drm/drm_print.c line 52
Quoted 6 lines
drivers/gpu/drm/drm_print.c +6
-0+/* +* __drm_debug_trace: Enable debug output in drm tracing instance. +* Bitmask of DRM_UT_x. See include/drm/drm_print.h for details. +*/ +unsigned int __drm_debug_trace; +EXPORT_SYMBOL (__drm_debug_trace);Hi!
Hi Pekka,
Thanks again for the feedback, I'm happy that we seem to be converging!
Quoted 9 lines
Might distributions perhaps want to set a default value for this via
Kconfig? Or could setting it via sysfs happen early enough to diagnose e.g. Plymouth problems?Or maybe there is nothing to see from early boot?
The general usefulness of this feature depends on whether people actually run with it enabled.
I had assumed that the cmdline argument would be sufficient for distros, is Kconfig preferable here? The module parameter has the advantage of being runtime configurable and is more in line with drm.debug. We can do either in CrOS, so I'm happy to go with crowd consensus.
Sean
Quoted 65 lines
Quoted 18 lines
drivers/gpu/drm/drm_print.c +15
-0+ +#define DEBUG_PARM_DESC (dst) \ +"Enable debug output to " dst ", where each bit enables a debug category.\n" \ +"\t\tBit 0 (0x01) will enable CORE messages (drm core code)\n" \ +"\t\tBit 1 (0x02) will enable DRIVER messages (drm controller code)\n" \ +"\t\tBit 2 (0x04) will enable KMS messages (modesetting code)\n" \ +"\t\tBit 3 (0x08) will enable PRIME messages (prime code)\n" \ +"\t\tBit 4 (0x10) will enable ATOMIC messages (atomic code)\n" \ +"\t\tBit 5 (0x20) will enable VBL messages (vblank code)\n" \ +"\t\tBit 7 (0x80) will enable LEASE messages (leasing code)\n" \ +"\t\tBit 8 (0x100) will enable DP messages (displayport code)" + +MODULE_PARM_DESC (debug,DEBUG_PARM_DESC ("syslog" ));module_param_named (debug, __drm_debug_syslog, int,0600 ); +MODULE_PARM_DESC (trace,DEBUG_PARM_DESC ("tracefs" )); +module_param_named (trace, __drm_debug_trace, int,0600 );...
Quoted 31 lines
drivers/gpu/drm/drm_print.c +31
-0+ +/** +* DOC: DRM Tracing +* +* *tl;dr* DRM tracing is a lightweight alternative to traditional DRM debug +* logging. +* +* While DRM logging is quite convenient when reproducing a specific issue, it +* doesn't help when something goes wrong unexpectedly. There are a couple +* reasons why one does not want to enable DRM logging at all times: +* +* 1. We don't want to overwhelm syslog with drm spam, others have to use it too +* 2. Console logging is slow +* +* DRM tracing aims to solve both these problems. +* +* To use DRM tracing, set the drm.trace module parameter (via cmdline or sysfs) +* to a DRM debug category mask (this is a bitmask of &drm_debug_category +* values): +* :: +* +* eg: echo 0x106 > /sys/module/drm/parameters/trace +* +* Once active, all log messages in the specified categories will be written to +* the DRM trace. Once at capacity, the trace will overwrite old messages with +* new ones. At any point, one can read the trace file to extract the previous N +* DRM messages: +* :: +* +* eg: cat /sys/kernel/tracing/instances/drm/trace +* drivers/gpu/drm/drm_print.c void drm_print_regset32(struct drm_printer *p, struct debugfs_regset32 *regset) · line 448
Quoted 10 lines
drivers/gpu/drm/drm_print.c +10
-0+* Considerations +* ************** +* The trace is subsystem wide, so if you have multiple devices active, they +* will be adding logs to the same trace. +* +* The contents of the DRM Trace are **not** considered UABI. **DO NOT depend on +* the values of these traces in your userspace.** These traces are intended for +* entertainment purposes only. The contents of these logs carry no warranty, +* expressed or implied. +*/ Sounds good to me!
This part is:
Quoted 5 lines
Acked-by Pekka Paalanen <[email protected]>Thanks,
pq
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v5 13/13] drm/print: Add tracefs support to the drm logging helpers
Quoted 42 lines · On Wed, 10 Jun 2020 09:29:37 -0400 Sean Paul <[email protected]> wrote:
Quoted 41 lines · Pekka Paalanen
Quoted 37 lines · On Mon, 8 Jun 2020 17:05:03 -0400 Sean Paul <[email protected]> wrote:
From: Sean Paul <[email protected]>
This patch adds a new module parameter called drm.trace which accepts the same mask as drm.debug. When a debug category is enabled, log messages will be put in a new tracefs instance called drm for consumption.
Using the new tracefs instance will allow distros to enable drm logging in production without impacting performance or spamming the system logs.
Cc Daniel Vetter <[email protected]>Cc David Airlie <[email protected]>Cc Jani Nikula <[email protected]>Cc Joonas Lahtinen <[email protected]>Cc Pekka Paalanen <[email protected]>Cc Rob Clark <[email protected]>Cc Steven Rostedt <[email protected]>Cc Thomas Zimmermann <[email protected]>Cc Ville Syrjälä <[email protected]>Cc Chris Wilson <[email protected]>Cc Steven Rostedt <[email protected]>Signed-off-by Sean Paul <[email protected]>Changes in v5:
-Re-write to usetrace_arrayand the tracefs instance support--- Documentation/gpu/drm-uapi.rst | 6 + drivers/gpu/drm/drm_drv.c | 3 + drivers/gpu/drm/drm_print.c | 209 ++++++++++++++++++++++++++++----- include/drm/drm_print.h | 63 ++++++++-- 4 files changed, 241 insertions(+), 40 deletions(-)
...
Quoted 34 lines
Quoted 15 lines
Quoted 15 lines
#include <
drm/drm.h>
#include <drm/drm_drv.h>drivers/gpu/drm/drm_print.c +0
-9@@ -43,17 +44,34 @@unsigned int __drm_debug_syslog;EXPORT_SYMBOL (__drm_debug_syslog); -MODULE_PARM_DESC (debug,"Enable debug output, where each bit enables a debug category.\n" -"\t\tBit 0 (0x01) will enable CORE messages (drm core code)\n" -"\t\tBit 1 (0x02) will enable DRIVER messages (drm controller code)\n" -"\t\tBit 2 (0x04) will enable KMS messages (modesetting code)\n" -"\t\tBit 3 (0x08) will enable PRIME messages (prime code)\n" -"\t\tBit 4 (0x10) will enable ATOMIC messages (atomic code)\n" -"\t\tBit 5 (0x20) will enable VBL messages (vblank code)\n" -"\t\tBit 7 (0x80) will enable LEASE messages (leasing code)\n" -"\t\tBit 8 (0x100) will enable DP messages (displayport code)" );Quoted 8 lines
drivers/gpu/drm/drm_print.c line 52
Quoted 6 lines
drivers/gpu/drm/drm_print.c +6
-0+/* +* __drm_debug_trace: Enable debug output in drm tracing instance. +* Bitmask of DRM_UT_x. See include/drm/drm_print.h for details. +*/ +unsigned int __drm_debug_trace; +EXPORT_SYMBOL (__drm_debug_trace);Hi!
Hi Pekka,
Thanks again for the feedback, I'm happy that we seem to be converging!Quoted 7 lines
Might distributions perhaps want to set a default value for this via
Kconfig? Or could setting it via sysfs happen early enough to diagnose e.g. Plymouth problems?Or maybe there is nothing to see from early boot?
Quoted 8 lines
Quoted 2 lines
The general usefulness of this feature depends on whether people actually run with it enabled.
I had assumed that the cmdline argument would be sufficient for distros, is
Kconfigpreferable here? The module parameter has the advantage of being runtime configurable and is more in line with drm.debug. We can do either in CrOS, so I'm happy to go with crowd consensus.
Hi,
I don't know. I did mean only *default* value for the module parameter via Kconfig. Right now your default value is implicit 0.
But that's just me thinking that making kernel command lines even longer would be somehow bad. Maybe it's fine?
Thanks,
pq
Sean Paul
reviewerQuoted 20 lines · Sean Paul
From: Sean Paul <[email protected]>
This series is the latest in my journey to create a lightweight, always-on "flight recorder" (name credit Weston) of drm logs. This incarnation uses a
trace_arrayto keep logs in memory exposed through tracefs. Users and distros can enable drm logs by using the drm.trace module parameter to choose the debug categories they are interested in.The set has ballooned a little bit since the last version. Reason being is that I decided to return true from
drm_debug_enabledif trace was enabled. This should make it easier and more seamless for driver developers to use the correct interface, but meant I needed to audit all uses ofdrm_debug_enabledanddrm_debug_printer.Out of all those calls, there are 3 situations which arose:
1- The logs should only go to syslog:
I've converted these to use thedrm_debug_syslogvariant of enable check/printer.
Quoted 9 lines
2- The logs should go to both syslog/trace, but were using
pr_debug():
I've converted these to use the proper drm logging helper. In cases which used adrm_printerI've had to use a newdrm_debug_category_printerto ensure they are routed correctly.3- The logs should go to both syslog/trace and are using drm logging fns:
These are untouched and should be routed to the appropriate place(s)
Friendly ping. Is this something upstream wants, or should we continue carrying it downstream?
Sean
Quoted 33 lines
Sean Paul (13):
drm/mipi_dbi: Convertpr_debugcalls toDRM_DEBUG_DRIVERdrm/sil164: Convertdev_printktodrm_dev_dbgdrm/i915/utils: Replacedev_printkwith drm helpers drm/msm/dpu: Replace definitions for dpu debug macros drm/print: rename drm_debug* to be more syslog-centric drm/amd: Gate i2c transaction logs ondrm_debug_syslogdrm/etnaviv: Change buffer dump checks to target syslog drm/nouveau: Change debug checks to specifically target syslog drm/i915: Change infoframe debug checks to specify syslog drm/print: Adddrm_debug_category_printer
drm/mst: Convert debug printers to debug category printers drm/i915: Use debug category printer for welcome message drm/print: Add tracefs support to the drm logging helpersDocumentation/gpu/drm-uapi.rst | 6 + drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c | 4 +- drivers/gpu/drm/drm_dp_mst_topology.c | 9 +- drivers/gpu/drm/drm_drv.c | 3 + drivers/gpu/drm/drm_mipi_dbi.c | 8 +- drivers/gpu/drm/drm_print.c | 228 ++++++++++++++++--- drivers/gpu/drm/etnaviv/etnaviv_buffer.c | 8 +- drivers/gpu/drm/i2c/sil164_drv.c | 12 +- drivers/gpu/drm/i915/display/intel_display.c | 4 +- drivers/gpu/drm/i915/i915_drv.c | 3 +- drivers/gpu/drm/i915/i915_utils.c | 5 +- drivers/gpu/drm/msm/disp/dpu1/dpu_kms.h | 20 +- drivers/gpu/drm/nouveau/dispnv50/disp.h | 4 +- drivers/gpu/drm/nouveau/nouveau_drv.h | 4 +- include/drm/drm_print.h | 96 +++++++- 15 files changed, 318 insertions(+), 96 deletions(-)
Quoted 3 lines
--
Sean Paul, Software Engineer, Google / Chromium OS
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel