Fix some issues with TRBE building as a module
cover Fix some issues with TRBE building as a module
The TRBE driver support is build as a module, we found some driver issues based on the patchset [1] and set CONFIG_CORESIGHT_TRBE=m.
1. TRBE driver potential sleep in atomic context when unregister device
2. Multiple free the platform data resource when rmmod coresight TRBE driver
[1] "coresight: trbe: Enable ACPI based devices"
https://lore.kernel.org/all/[email protected]/
Junhao He (2):
coresight: trbe: Fix TRBE potential sleep in atomic context
coresight: core: Fix multiple free TRBE platform data resource
drivers/hwtracing/coresight/coresight-core.c | 7 ++-- drivers/hwtracing/coresight/coresight-trbe.c | 35 +++++++++++--------- 2 files changed, 24 insertions(+), 18 deletions(-)
Signature · 7 lines
2.33.0 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 1/2 coresight: trbe: Fix TRBE potential sleep in atomic context
smp_call_function_single() will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will execute arm_trbe_remove_coresight_cpu() call request in the interrupt handler.
According to the device_unregister() stack information, if other process is useing the device, the down_write() may sleep, and trigger deadlocks or unexpected errors.
arm_trbe_remove_coresight_cpu
coresight_unregister
device_unregister
device_del
kobject_del
__kobject_del
sysfs_remove_dir
kernfs_remove
down_write ---------> it may sleepAdd a helper arm_trbe_disable_cpu() to disable TRBE precpu irq and reset per TRBE.
Simply call arm_trbe_remove_coresight_cpu() directly without useing the smp_call_function_single(), which is the same as registering the TRBE coresight device.
--- drivers/hwtracing/coresight/coresight-trbe.c | 35 +++++++++++--------- 1 file changed, 20 insertions(+), 15 deletions(-)
drivers/hwtracing/coresight/coresight-trbe.c +20 -15
@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info)enable_percpu_irq (drvdata->irq,IRQ_TYPE_NONE ); } +static void arm_trbe_disable_cpu (void *info) +{ +struct trbe_drvdata *drvdata = info; +struct trbe_cpudata *cpudata =this_cpu_ptr (drvdata->cpudata); + +disable_percpu_irq (drvdata->irq); +trbe_reset_local (cpudata); + cpudata->drvdata =NULL ; +} + +static void arm_trbe_register_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) {struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);
@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info)cpumask_clear_cpu (cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu (void *info) +static void arm_trbe_remove_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) { -int cpu =smp_processor_id (); -struct trbe_drvdata *drvdata = info; -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);struct coresight_device *trbe_csdev =coresight_get_percpu_sink (cpu); -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata);if (trbe_csdev) {coresight_unregister (trbe_csdev); - cpudata->drvdata =NULL ;coresight_set_percpu_sink (cpu,NULL ); } }
@@ -1366,8 +1371,12 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) {int cpu; -for_each_cpu (cpu, &drvdata->supported_cpus) -smp_call_function_single (cpu, arm_trbe_remove_coresight_cpu, drvdata,1 ); +for_each_cpu (cpu, &drvdata->supported_cpus) { +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +smp_call_function_single (cpu, arm_trbe_disable_cpu, drvdata,1 ); +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_remove_coresight_cpu (drvdata, cpu); + }free_percpu (drvdata->cpudata);return 0 ; }
@@ -1406,12 +1415,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) { struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node); -if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) { -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu); - -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata); - } +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu (drvdata);return 0 ; }
Signature · 7 lines
2.33.0 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
Hi Junhao
Quoted 62 lines · Junhao He
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.arm_trbe_remove_coresight_cpu coresight_unregister device_unregister device_del kobject_del __kobject_del sysfs_remove_dir kernfs_remove down_write ---------> it may sleepAdd a helper
arm_trbe_disable_cpu()to disable TRBE precpu irq and reset per TRBE.
Simply callarm_trbe_remove_coresight_cpu()directly without useing thesmp_call_function_single(), which is the same as registering the TRBE coresight device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Signed-off-by Junhao He <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 35 +++++++++++--------- 1 file changed, 20 insertions(+), 15 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +12
-2@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info)enable_percpu_irq (drvdata->irq,IRQ_TYPE_NONE ); } +static void arm_trbe_disable_cpu (void *info) +{ +struct trbe_drvdata *drvdata = info; +struct trbe_cpudata *cpudata =this_cpu_ptr (drvdata->cpudata); + +disable_percpu_irq (drvdata->irq); +trbe_reset_local (cpudata); + cpudata->drvdata =NULL ; +} + +static void arm_trbe_register_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) {struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info)cpumask_clear_cpu (cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu (void *info) +static void arm_trbe_remove_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) { -int cpu =smp_processor_id ();
drivers/hwtracing/coresight/coresight-trbe.c static void arm_trbe_probe_cpu(void *info) · line 1346
Quoted 10 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-5-struct trbe_drvdata *drvdata = info; -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);struct coresight_device *trbe_csdev =coresight_get_percpu_sink (cpu); -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata);if (trbe_csdev) {coresight_unregister (trbe_csdev); - cpudata->drvdata =NULL ;coresight_set_percpu_sink (cpu,NULL );
I am a bit concerned about "resetting" the sink from a different CPU. Could we instead, schedule a delayed work to unregister the trbe_csdev?
Suzuki
Quoted 22 lines
drivers/hwtracing/coresight/coresight-trbe.c +6
-3} }@@ -1366,8 +1371,12 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) {int cpu; -for_each_cpu (cpu, &drvdata->supported_cpus) -smp_call_function_single (cpu, arm_trbe_remove_coresight_cpu, drvdata,1 ); +for_each_cpu (cpu, &drvdata->supported_cpus) { +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +smp_call_function_single (cpu, arm_trbe_disable_cpu, drvdata,1 ); +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_remove_coresight_cpu (drvdata, cpu); + }free_percpu (drvdata->cpudata);return 0 ; }@@ -1406,12 +1415,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) { struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node); -if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) {
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) · line 1422
Quoted 10 lines
drivers/hwtracing/coresight/coresight-trbe.c +2
-5-struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu); - -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata); - } +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu (drvdata);return 0 ; }
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
Hi Suzuki
Quoted 70 lines · Suzuki K Poulose
Hi Junhao
Quoted 67 lines · Junhao He
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.arm_trbe_remove_coresight_cpu coresight_unregister device_unregister device_del kobject_del __kobject_del sysfs_remove_dir kernfs_remove down_write ---------> it may sleepAdd a helper
arm_trbe_disable_cpu()to disable TRBE precpu irq and reset per TRBE.
Simply callarm_trbe_remove_coresight_cpu()directly without useing thesmp_call_function_single(), which is the same as registering the TRBE coresight device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Signed-off-by Junhao He <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 35 +++++++++++--------- 1 file changed, 20 insertions(+), 15 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +11
-3@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info)enable_percpu_irq (drvdata->irq,IRQ_TYPE_NONE ); } +static void arm_trbe_disable_cpu(void *info) +{ +struct trbe_drvdata *drvdata = info; +struct trbe_cpudata *cpudata =this_cpu_ptr (drvdata->cpudata); + +disable_percpu_irq (drvdata->irq); +trbe_reset_local (cpudata); + cpudata->drvdata =NULL ; +} + +static void arm_trbe_register_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) {struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info)cpumask_clear_cpu (cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu(void *info) +static void arm_trbe_remove_coresight_cpu(struct trbe_drvdata *drvdata, int cpu) { -int cpu =smp_processor_id (); -struct trbe_drvdata *drvdata = info; -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);struct coresight_device *trbe_csdev =coresight_get_percpu_sink (cpu); -disable_percpu_irq (drvdata->irq);
Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c static void arm_trbe_probe_cpu(void *info) · line 1346
Quoted 5 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-2-trbe_reset_local (cpudata);if (trbe_csdev) {coresight_unregister (trbe_csdev); - cpudata->drvdata =NULL ;coresight_set_percpu_sink (cpu,NULL );I am a bit concerned about "resetting" the sink from a different CPU. Could we instead, schedule a delayed work to unregister the
trbe_csdev?
Yes, I will try to do that.
Sorry for my following questions.
As you mean, do we need to take the same care when setting the percpu sink in the register trbe_csdev ?
Best regards,
Junhao.
Quoted 32 lines
Quoted 32 lines
drivers/hwtracing/coresight/coresight-trbe.c +6
-5} }@@ -1366,8 +1371,12 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) {int cpu; -for_each_cpu (cpu, &drvdata->supported_cpus) -smp_call_function_single (cpu, arm_trbe_remove_coresight_cpu, drvdata, 1); +for_each_cpu (cpu, &drvdata->supported_cpus) { +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +smp_call_function_single (cpu, arm_trbe_disable_cpu, drvdata,1 ); +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_remove_coresight_cpu (drvdata, cpu); + }free_percpu (drvdata->cpudata);return 0 ; }@@ -1406,12 +1415,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) { struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node); - if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) { -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu); - -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata);
Quoted 9 lines
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) · line 1421
Quoted 5 lines
drivers/hwtracing/coresight/coresight-trbe.c +2
-1- } +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu (drvdata);return 0 ; }.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
Quoted 81 lines · hejunhao
Hi Suzuki
Quoted 68 lines · Suzuki K Poulose
Hi Junhao
Quoted 65 lines · Junhao He
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.
arm_trbe_remove_coresight_cpucoresight_unregister
device_unregister
device_del
kobject_del
__kobject_del
sysfs_remove_dir
kernfs_remove
down_write---------> it may sleepAdd a helper
arm_trbe_disable_cpu()to disable TRBE precpu irq and reset per TRBE.
Simply callarm_trbe_remove_coresight_cpu()directly without useing thesmp_call_function_single(), which is the same as registering the TRBE coresight device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Signed-off-by Junhao He <[email protected]>---
drivers/hwtracing/coresight/coresight-trbe.c| 35 +++++++++++---------1 file changed, 20 insertions(+), 15 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +11
-3@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info) enable_percpu_irq(drvdata->irq, IRQ_TYPE_NONE); } +static void arm_trbe_disable_cpu(void *info) +{ + struct trbe_drvdata *drvdata = info; + struct trbe_cpudata *cpudata = this_cpu_ptr(drvdata->cpudata); + + disable_percpu_irq(drvdata->irq); + trbe_reset_local(cpudata); + cpudata->drvdata = NULL; +} + + static void arm_trbe_register_coresight_cpu(struct trbe_drvdata *drvdata, int cpu) { struct trbe_cpudata *cpudata = per_cpu_ptr(drvdata->cpudata, cpu);@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info) cpumask_clear_cpu(cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu(void *info) +static void arm_trbe_remove_coresight_cpu(struct trbe_drvdata *drvdata, int cpu) { - int cpu = smp_processor_id(); - struct trbe_drvdata *drvdata = info; - struct trbe_cpudata *cpudata = per_cpu_ptr(drvdata->cpudata, cpu); struct coresight_device *trbe_csdev = coresight_get_percpu_sink(cpu); - disable_percpu_irq(drvdata->irq);Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c static void arm_trbe_probe_cpu(void *info) · line 1346
Quoted 5 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-2- trbe_reset_local(cpudata); if (trbe_csdev) { coresight_unregister(trbe_csdev); - cpudata->drvdata = NULL; coresight_set_percpu_sink(cpu, NULL);I am a bit concerned about "resetting" the sink from a different CPU. Could we instead, schedule a delayed work to unregister the
trbe_csdev?
Quoted 4 lines
Yes, I will try to do that.
Sorry for my following questions.
As you mean, do we need to take the same care when setting the percpu sink in the registertrbe_csdev?
Apologies, having taken another look, we set the percpu_sink for a cpu outside smp_call_function(). So, I think your patch is fine.
Quoted 6 lines
Best regards,
Junhao.Quoted 2 lines
Quoted 15 lines
Quoted 7 lines
Quoted 7 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-0}}
drivers/hwtracing/coresight/coresight-trbe.c +0
-0@@ -1366,8 +1371,12 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) { int cpu; - for_each_cpu(cpu, &drvdata->supported_cpus)Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) · line 1378
Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c +5
-1- smp_call_function_single(cpu, arm_trbe_remove_coresight_cpu, drvdata, 1); + for_each_cpu(cpu, &drvdata->supported_cpus) { + if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) + smp_call_function_single(cpu, arm_trbe_disable_cpu, drvdata,1 ); + if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) + arm_trbe_remove_coresight_cpu(drvdata, cpu);
Do we need to test the cpu here in both places ? We already check that in the loop entry. The reason why we repeat the check during the probe, is to skip any CPUs that may have a TRBE not accessible.
Suzuki
Quoted 15 lines
Quoted 15 lines
Quoted 15 lines
drivers/hwtracing/coresight/coresight-trbe.c +1
-0+ }
free_percpu(drvdata->cpudata);
return 0;
}drivers/hwtracing/coresight/coresight-trbe.c +0
-5@@ -1406,12 +1415,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) { struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node); - if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) { - struct trbe_cpudata *cpudata = per_cpu_ptr(drvdata->cpudata, cpu); - - disable_percpu_irq(drvdata->irq); - trbe_reset_local(cpudata); - }
Quoted 9 lines
Quoted 8 lines
Quoted 4 lines
drivers/hwtracing/coresight/coresight-trbe.c +2
-0+ if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) + arm_trbe_disable_cpu(drvdata); return 0;}
.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
Hi Suzuki
Quoted 122 lines · Suzuki K Poulose
Quoted 86 lines · hejunhao
Hi Suzuki
Quoted 73 lines · Suzuki K Poulose
Hi Junhao
Quoted 70 lines · Junhao He
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.arm_trbe_remove_coresight_cpu coresight_unregister device_unregister device_del kobject_del __kobject_del sysfs_remove_dir kernfs_remove down_write ---------> it may sleepAdd a helper
arm_trbe_disable_cpu()to disable TRBE precpu irq and reset per TRBE.
Simply callarm_trbe_remove_coresight_cpu()directly without useing thesmp_call_function_single(), which is the same as registering the TRBE coresight device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Signed-off-by Junhao He <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 35 +++++++++++--------- 1 file changed, 20 insertions(+), 15 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +10
-0@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info)enable_percpu_irq (drvdata->irq,IRQ_TYPE_NONE ); } +static void arm_trbe_disable_cpu(void *info) +{ +struct trbe_drvdata *drvdata = info; +struct trbe_cpudata *cpudata =this_cpu_ptr (drvdata->cpudata); + +disable_percpu_irq (drvdata->irq); +trbe_reset_local (cpudata); + cpudata->drvdata =NULL ; +} + +static void arm_trbe_register_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) { struct trbe_cpudata *cpudata = per_cpu_ptr(drvdata->cpudata,cpu);
drivers/hwtracing/coresight/coresight-trbe.c +1
-3@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info)cpumask_clear_cpu (cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu(void *info) +static void arm_trbe_remove_coresight_cpu(struct trbe_drvdata *drvdata, int cpu) { -int cpu =smp_processor_id (); -struct trbe_drvdata *drvdata = info; -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);struct coresight_device *trbe_csdev =coresight_get_percpu_sink (cpu); -disable_percpu_irq (drvdata->irq);Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c static void arm_trbe_probe_cpu(void *info) · line 1346
Quoted 5 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-2-trbe_reset_local (cpudata);if (trbe_csdev) {coresight_unregister (trbe_csdev); - cpudata->drvdata =NULL ;coresight_set_percpu_sink (cpu,NULL );I am a bit concerned about "resetting" the sink from a different CPU. Could we instead, schedule a delayed work to unregister the
trbe_csdev?Quoted 4 lines
Yes, I will try to do that.
Sorry for my following questions.
As you mean, do we need to take the same care when setting the percpu sink in the registertrbe_csdev?Apologies, having taken another look, we set the
percpu_sinkfor a cpu outsidesmp_call_function(). So, I think your patch is fine.Quoted 6 lines
Best regards,
Junhao.Quoted 2 lines
Quoted 15 lines
Quoted 7 lines
Quoted 7 lines
}}
drivers/hwtracing/coresight/coresight-trbe.c +0
-0@@ -1366,8 +1371,12 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) {int cpu; -for_each_cpu (cpu, &drvdata->supported_cpus)Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) · line 1378
Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c +5
-1-smp_call_function_single (cpu, arm_trbe_remove_coresight_cpu, drvdata, 1); +for_each_cpu (cpu, &drvdata->supported_cpus) { +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +smp_call_function_single (cpu, arm_trbe_disable_cpu, drvdata,1 ); +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_remove_coresight_cpu (drvdata, cpu);Do we need to test the cpu here in both places ? We already check that in the loop entry. The reason why we repeat the check during the probe, is to skip any CPUs that may have a TRBE not accessible.
Quoted 2 lines
Suzuki
Ok, Will fix in next version.
Best regards,
Junhao.
Quoted 28 lines
Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
drivers/hwtracing/coresight/coresight-trbe.c +1
-5+ }free_percpu (drvdata->cpudata);return 0 ; }@@ -1406,12 +1415,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) { struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node); - if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) { -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu); - -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata); - }Quoted 9 lines
Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) · line 1421
Quoted 4 lines
drivers/hwtracing/coresight/coresight-trbe.c +2
-0+if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu (drvdata);return 0 ; }.
Quoted 2 lines
.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 2/2 coresight: core: Fix multiple free TRBE platform data resource
In reply to Junhao HeCurrent the TRBE driver supports matching TRBE platform device through id_table. The ACPI created a dummy TRBE platform device inside drivers/perf/arm_pmu_acpi.c. So the TRBE platform driver will probe only once and allocate just one TRBE platform data resource.
If the system supports the TRBE feature, Each CPU in the systems can have at least one TRBE present, and the coresight_unregister gets called multiple times, once for each of them.
Therefore, when unregister TRBE coresight devices, the TRBE platform data resource will multiple free in function coresight_unregister.
root@localhost:# insmod coresight-trbe.ko root@localhost:# rmmod coresight-trbe.ko [ 423.455932] ------------[ cut here ]------------ [ 423.461987] WARNING: CPU: 1 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 [ 423.483821] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G O 6.5.0-rc4+ #1 [ 423.505842] pstate: 614000c9 (nZCv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--) ... [ 423.601301] Call trace: [ 423.604202] devm_kfree+0x88/0x98 [ 423.608369] coresight_release_platform_data+0xb8/0xe0 [coresight] [ 423.616589] coresight_unregister+0x120/0x170 [coresight] [ 423.623533] arm_trbe_remove_coresight_cpu+0x70/0xa0 [coresight_trbe] [ 423.631082] __flush_smp_call_function_queue+0x1e4/0x4e0 [ 423.637471] generic_smp_call_function_single_interrupt+0x1c/0x30 [ 423.644796] ipi_handler+0x90/0x278 [ 423.648992] handle_percpu_devid_irq+0x90/0x250 [ 423.654636] generic_handle_domain_irq+0x34/0x58 [ 423.659786] gic_handle_irq+0x12c/0x270 [ 423.664039] call_on_irq_stack+0x24/0x30 [ 423.668452] do_interrupt_handler+0x88/0x98 [ 423.673027] el1_interrupt+0x48/0xe8 [ 423.677413] el1h_64_irq_handler+0x18/0x28 [ 423.681781] el1h_64_irq+0x78/0x80 [ 423.685550] default_idle_call+0x5c/0x180 [ 423.689855] do_idle+0x25c/0x2c0 [ 423.694196] cpu_startup_entry+0x2c/0x40 [ 423.698373] secondary_start_kernel+0x144/0x188 [ 423.703920] __secondary_switched+0xb8/0xc0 [ 423.708972] ---[ end trace 0000000000000000 ]--- [ 423.729209] ------------[ cut here ]------------ ... [ 423.735217] WARNING: CPU: 2 PID: 40 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ... [ 424.012385] WARNING: CPU: 3 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ...
This patch does the following:
1.TRBE coresight devices do not need regular connections information, We
can free connections resource when the nr_conns is valid.
2.And we can ignore the free platform data resource, it will be
automatically free in platform_driver_unregister().
--- drivers/hwtracing/coresight/coresight-core.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/hwtracing/coresight/coresight-core.c +4 -3
@@ -1555,9 +1555,10 @@ void coresight_release_platform_data(struct coresight_device *csdev, conns[i]->dest_fwnode =NULL ;devm_kfree (dev, conns[i]); } -devm_kfree (dev, pdata->out_conns); -devm_kfree (dev, pdata->in_conns); -devm_kfree (dev, pdata); +if (pdata->nr_outconns) +devm_kfree (dev, pdata->out_conns); +if (pdata->nr_inconns) +devm_kfree (dev, pdata->in_conns);if (csdev)coresight_remove_conns_sysfs_group (csdev); }
Signature · 7 lines
2.33.0 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 2/2] coresight: core: Fix multiple free TRBE platform data resource
+ James Clark
Quoted 47 lines · Junhao He
Current the TRBE driver supports matching TRBE platform device through
id_table. The ACPI created a dummy TRBE platform device insidedrivers/perf/arm_pmu_acpi.c. So the TRBE platform driver will probe only once and allocate just one TRBE platform data resource.If the system supports the TRBE feature, Each CPU in the systems can have at least one TRBE present, and the
coresight_unregistergets called multiple times, once for each of them.
Therefore, when unregister TRBE coresight devices, the TRBE platform data resource will multiple free in functioncoresight_unregister.root@localhost:# insmod coresight-trbe.ko root@localhost:# rmmod coresight-trbe.ko [ 423.455932] ------------[ cut here ]------------ [ 423.461987] WARNING: CPU: 1 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 [ 423.483821] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G O 6.5.0-rc4+ #1 [ 423.505842] pstate: 614000c9 (nZCv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--) ... [ 423.601301] Call trace: [ 423.604202] devm_kfree+0x88/0x98 [ 423.608369] coresight_release_platform_data+0xb8/0xe0 [coresight] [ 423.616589] coresight_unregister+0x120/0x170 [coresight] [ 423.623533] arm_trbe_remove_coresight_cpu+0x70/0xa0 [coresight_trbe] [ 423.631082] __flush_smp_call_function_queue+0x1e4/0x4e0 [ 423.637471] generic_smp_call_function_single_interrupt+0x1c/0x30 [ 423.644796] ipi_handler+0x90/0x278 [ 423.648992] handle_percpu_devid_irq+0x90/0x250 [ 423.654636] generic_handle_domain_irq+0x34/0x58 [ 423.659786] gic_handle_irq+0x12c/0x270 [ 423.664039] call_on_irq_stack+0x24/0x30 [ 423.668452] do_interrupt_handler+0x88/0x98 [ 423.673027] el1_interrupt+0x48/0xe8 [ 423.677413] el1h_64_irq_handler+0x18/0x28 [ 423.681781] el1h_64_irq+0x78/0x80 [ 423.685550] default_idle_call+0x5c/0x180 [ 423.689855] do_idle+0x25c/0x2c0 [ 423.694196] cpu_startup_entry+0x2c/0x40 [ 423.698373] secondary_start_kernel+0x144/0x188 [ 423.703920] __secondary_switched+0xb8/0xc0 [ 423.708972] ---[ end trace 0000000000000000 ]--- [ 423.729209] ------------[ cut here ]------------ ... [ 423.735217] WARNING: CPU: 2 PID: 40 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ... [ 424.012385] WARNING: CPU: 3 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ...
Quoted 5 lines
This patch does the following:
1.TRBE coresight devices do not need regular connections information, We
can free connections resource when thenr_connsis valid.
2.And we can ignore the free platform data resource, it will be
automatically free inplatform_driver_unregister().
Do we need a Fixes tag here ?
Quoted 11 lines
Signed-off-by Junhao He <[email protected]>--- drivers/hwtracing/coresight/coresight-core.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)drivers/hwtracing/coresight/coresight-core.c +0
-0
drivers/hwtracing/coresight/coresight-core.c void coresight_release_platform_data(struct coresight_device *csdev, · line 1561
Quoted 10 lines
drivers/hwtracing/coresight/coresight-core.c +4
-3@@ -1555,9 +1555,10 @@ void coresight_release_platform_data(struct coresight_device *csdev, conns[i]->dest_fwnode =NULL ;devm_kfree (dev, conns[i]); } -devm_kfree (dev, pdata->out_conns); -devm_kfree (dev, pdata->in_conns); -devm_kfree (dev, pdata); +if (pdata->nr_outconns) +devm_kfree (dev, pdata->out_conns); +if (pdata->nr_inconns) +devm_kfree (dev, pdata->in_conns);
These allocations are made on the parent device and that may never get unregistered (e.g., AMBA device, platform device, stay forever, even when the "coresight" modules are unloaded). Thus the memory will be left unused, literally leaking. This specific devm_kfree() was added to fix that. May be we should fix this in the TRBE driver to use separate pdata for the TRBE device instances.
Suzuki
Quoted 3 lines
drivers/hwtracing/coresight/coresight-core.c +0
-0if (csdev)coresight_remove_conns_sysfs_group (csdev); }
Re: [PATCH 2/2] coresight: core: Fix multiple free TRBE platform data resource
Hi, Suzuki
Quoted 50 lines · Suzuki K Poulose
+ James Clark
Quoted 47 lines · Junhao He
Current the TRBE driver supports matching TRBE platform device through
id_table. The ACPI created a dummy TRBE platform device insidedrivers/perf/arm_pmu_acpi.c. So the TRBE platform driver will probe only once and allocate just one TRBE platform data resource.If the system supports the TRBE feature, Each CPU in the systems can have at least one TRBE present, and the
coresight_unregistergets called multiple times, once for each of them.
Therefore, when unregister TRBE coresight devices, the TRBE platform data resource will multiple free in functioncoresight_unregister.root@localhost:# insmod coresight-trbe.ko root@localhost:# rmmod coresight-trbe.ko [ 423.455932] ------------[ cut here ]------------ [ 423.461987] WARNING: CPU: 1 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 [ 423.483821] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G O 6.5.0-rc4+ #1 [ 423.505842] pstate: 614000c9 (nZCv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--) ... [ 423.601301] Call trace: [ 423.604202] devm_kfree+0x88/0x98 [ 423.608369] coresight_release_platform_data+0xb8/0xe0 [coresight] [ 423.616589] coresight_unregister+0x120/0x170 [coresight] [ 423.623533] arm_trbe_remove_coresight_cpu+0x70/0xa0 [coresight_trbe] [ 423.631082] __flush_smp_call_function_queue+0x1e4/0x4e0 [ 423.637471] generic_smp_call_function_single_interrupt+0x1c/0x30 [ 423.644796] ipi_handler+0x90/0x278 [ 423.648992] handle_percpu_devid_irq+0x90/0x250 [ 423.654636] generic_handle_domain_irq+0x34/0x58 [ 423.659786] gic_handle_irq+0x12c/0x270 [ 423.664039] call_on_irq_stack+0x24/0x30 [ 423.668452] do_interrupt_handler+0x88/0x98 [ 423.673027] el1_interrupt+0x48/0xe8 [ 423.677413] el1h_64_irq_handler+0x18/0x28 [ 423.681781] el1h_64_irq+0x78/0x80 [ 423.685550] default_idle_call+0x5c/0x180 [ 423.689855] do_idle+0x25c/0x2c0 [ 423.694196] cpu_startup_entry+0x2c/0x40 [ 423.698373] secondary_start_kernel+0x144/0x188 [ 423.703920] __secondary_switched+0xb8/0xc0 [ 423.708972] ---[ end trace 0000000000000000 ]--- [ 423.729209] ------------[ cut here ]------------ ... [ 423.735217] WARNING: CPU: 2 PID: 40 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ... [ 424.012385] WARNING: CPU: 3 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ...
Quoted 7 lines
Quoted 5 lines
This patch does the following:
1.TRBE coresight devices do not need regular connections information, We
can free connections resource when thenr_connsis valid.
2.And we can ignore the free platform data resource, it will be
automatically free inplatform_driver_unregister().Do we need a Fixes tag here ?
Yes, I will do that.
Quoted 31 lines
Quoted 12 lines
Signed-off-by Junhao He <[email protected]>--- drivers/hwtracing/coresight/coresight-core.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)drivers/hwtracing/coresight/coresight-core.c +0
-0@@ -1555,9 +1555,10 @@ void coresight_release_platform_data(struct coresight_device *csdev,drivers/hwtracing/coresight/coresight-core.c void coresight_release_platform_data(struct · line 1562
Quoted 10 lines
drivers/hwtracing/coresight/coresight-core.c +4
-3conns[i]->dest_fwnode =NULL ;devm_kfree (dev, conns[i]); } -devm_kfree (dev, pdata->out_conns); -devm_kfree (dev, pdata->in_conns); -devm_kfree (dev, pdata); +if (pdata->nr_outconns) +devm_kfree (dev, pdata->out_conns); +if (pdata->nr_inconns) +devm_kfree (dev, pdata->in_conns);These allocations are made on the parent device and that may never get unregistered (e.g., AMBA device, platform device, stay forever, even when the "coresight" modules are unloaded). Thus the memory will be left unused, literally leaking. This specific
devm_kfree()was added to fix that. May be we should fix this in the TRBE driver to use separate pdata for the TRBE device instances.
Quoted 1 line
Suzuki
If we fix this with minimal changes, I think it is possible to add a check and not free pdata if it is TRBE?
if (csdev->subtype.sink_subtype !=
CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM)
devm_kfree(dev, pdata);Then free pdata in the end of arm_trbe_remove_coresight().
Quoted 8 lines
Quoted 4 lines
drivers/hwtracing/coresight/coresight-core.c +0
-0if (csdev)coresight_remove_conns_sysfs_group (csdev);}.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 2/2] coresight: core: Fix multiple free TRBE platform data resource
Quoted 97 lines · hejunhao
Hi, Suzuki
Quoted 50 lines · Suzuki K Poulose
+ James Clark
Quoted 47 lines · Junhao He
Current the TRBE driver supports matching TRBE platform device through
id_table. The ACPI created a dummy TRBE platform device insidedrivers/perf/arm_pmu_acpi.c. So the TRBE platform driver will probe only once and allocate just one TRBE platform data resource.If the system supports the TRBE feature, Each CPU in the systems can have at least one TRBE present, and the
coresight_unregistergets called multiple times, once for each of them.
Therefore, when unregister TRBE coresight devices, the TRBE platform data resource will multiple free in functioncoresight_unregister.root@localhost:# insmod coresight-trbe.ko root@localhost:# rmmod coresight-trbe.ko [ 423.455932] ------------[ cut here ]------------ [ 423.461987] WARNING: CPU: 1 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 [ 423.483821] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G O 6.5.0-rc4+ #1 [ 423.505842] pstate: 614000c9 (nZCv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--) ... [ 423.601301] Call trace: [ 423.604202] devm_kfree+0x88/0x98 [ 423.608369] coresight_release_platform_data+0xb8/0xe0 [coresight] [ 423.616589] coresight_unregister+0x120/0x170 [coresight] [ 423.623533] arm_trbe_remove_coresight_cpu+0x70/0xa0 [coresight_trbe] [ 423.631082] __flush_smp_call_function_queue+0x1e4/0x4e0 [ 423.637471] generic_smp_call_function_single_interrupt+0x1c/0x30 [ 423.644796] ipi_handler+0x90/0x278 [ 423.648992] handle_percpu_devid_irq+0x90/0x250 [ 423.654636] generic_handle_domain_irq+0x34/0x58 [ 423.659786] gic_handle_irq+0x12c/0x270 [ 423.664039] call_on_irq_stack+0x24/0x30 [ 423.668452] do_interrupt_handler+0x88/0x98 [ 423.673027] el1_interrupt+0x48/0xe8 [ 423.677413] el1h_64_irq_handler+0x18/0x28 [ 423.681781] el1h_64_irq+0x78/0x80 [ 423.685550] default_idle_call+0x5c/0x180 [ 423.689855] do_idle+0x25c/0x2c0 [ 423.694196] cpu_startup_entry+0x2c/0x40 [ 423.698373] secondary_start_kernel+0x144/0x188 [ 423.703920] __secondary_switched+0xb8/0xc0 [ 423.708972] ---[ end trace 0000000000000000 ]--- [ 423.729209] ------------[ cut here ]------------ ... [ 423.735217] WARNING: CPU: 2 PID: 40 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ... [ 424.012385] WARNING: CPU: 3 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ...Quoted 7 lines
Quoted 5 lines
This patch does the following:
1.TRBE coresight devices do not need regular connections information, We can free connections resource when thenr_connsis valid. 2.And we can ignore the free platform data resource, it will be automatically free inplatform_driver_unregister().Do we need a Fixes tag here ?
Yes, I will do that.
Quoted 31 lines
Quoted 12 lines
Signed-off-by Junhao He <[email protected]>---
drivers/hwtracing/coresight/coresight-core.c| 7 ++++---1 file changed, 4 insertions(+), 3 deletions(-)drivers/hwtracing/coresight/coresight-core.c +0
-0@@ -1555,9 +1555,10 @@ void coresight_release_platform_data(struct coresight_device *csdev,drivers/hwtracing/coresight/coresight-core.c void coresight_release_platform_data(struct · line 1562
Quoted 10 lines
drivers/hwtracing/coresight/coresight-core.c +4
-3conns[i]->dest_fwnode = NULL; devm_kfree(dev, conns[i]); } - devm_kfree(dev, pdata->out_conns); - devm_kfree(dev, pdata->in_conns); - devm_kfree(dev, pdata); + if (pdata->nr_outconns) + devm_kfree(dev, pdata->out_conns); + if (pdata->nr_inconns) + devm_kfree(dev, pdata->in_conns);These allocations are made on the parent device and that may never get unregistered (e.g., AMBA device, platform device, stay forever, even when the "coresight" modules are unloaded). Thus the memory will be left unused, literally leaking. This specific
devm_kfree()was added to fix that. May be we should fix this in the TRBE driver to use separate pdata for the TRBE device instances.Quoted 1 line
Suzuki
Quoted 8 lines
If we fix this with minimal changes, I think it is possible to add a check and not free pdata if it is TRBE?
if (
csdev->subtype.sink_subtype!=CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM)
devm_kfree(dev, pdata);Then free pdata in the end of
arm_trbe_remove_coresight().
It is much nicer to do something like :
--8>--
coresight: trbe: Allocate platform data per device
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit 4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata per coresight_device.
--- drivers/hwtracing/coresight/coresight-trbe.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-)
drivers/hwtracing/coresight/coresight-trbe.c +8 -7
@@ -1255,10 +1255,17 @@ static void arm_trbe_register_coresight_cpu(struct trbe_drvdata *drvdata, int cpif (!desc.name) goto cpu_clear; +/* +* TRBE doesn't have any connections described via firmware. Instead +* we register the trbe instance as per-cpu sink. +*/ + desc.pdata =coresight_get_platform_data (dev); +if (IS_ERR (desc.pdata)) + goto cpu_clear; + desc.type =CORESIGHT_DEV_TYPE_SINK ; desc.subtype.sink_subtype =CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM ; desc.ops = &arm_trbe_cs_ops; - desc.pdata =dev_get_platdata (dev); desc.groups = arm_trbe_groups; desc.dev = dev; trbe_csdev =coresight_register (&desc);
@@ -1482,7 +1489,6 @@ static void arm_trbe_remove_irq(struct trbe_drvdata *drvdata)static int arm_trbe_device_probe (struct platform_device *pdev) { -struct coresight_platform_data *pdata;struct trbe_drvdata *drvdata;struct device *dev = &pdev->dev;int ret;
@@ -1497,12 +1503,7 @@ static int arm_trbe_device_probe(struct platform_device *pdev)if (!drvdata)return -ENOMEM ; - pdata =coresight_get_platform_data (dev); -if (IS_ERR (pdata)) -return PTR_ERR (pdata); -dev_set_drvdata (dev, drvdata); - dev->platform_data = pdata; drvdata->pdev = pdev; ret =arm_trbe_probe_irq (pdev, drvdata);if (ret)
Signature · 1 line
2.34.1
Quoted 10 lines
Quoted 9 lines
Quoted 5 lines
drivers/hwtracing/coresight/coresight-core.c +0
-0if (csdev) coresight_remove_conns_sysfs_group(csdev);}
.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 2/2] coresight: core: Fix multiple free TRBE platform data resource
Quoted 134 lines · Suzuki K Poulose
Quoted 97 lines · hejunhao
Hi, Suzuki
Quoted 50 lines · Suzuki K Poulose
+ James Clark
Quoted 47 lines · Junhao He
Current the TRBE driver supports matching TRBE platform device through
id_table. The ACPI created a dummy TRBE platform device insidedrivers/perf/arm_pmu_acpi.c. So the TRBE platform driver will probe only once and allocate just one TRBE platform data resource.If the system supports the TRBE feature, Each CPU in the systems can have at least one TRBE present, and the
coresight_unregistergets called multiple times, once for each of them.
Therefore, when unregister TRBE coresight devices, the TRBE platform data resource will multiple free in functioncoresight_unregister.root@localhost:# insmod coresight-trbe.ko root@localhost:# rmmod coresight-trbe.ko [ 423.455932] ------------[ cut here ]------------ [ 423.461987] WARNING: CPU: 1 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 [ 423.483821] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G O 6.5.0-rc4+ #1 [ 423.505842] pstate: 614000c9 (nZCv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--) ... [ 423.601301] Call trace: [ 423.604202] devm_kfree+0x88/0x98 [ 423.608369] coresight_release_platform_data+0xb8/0xe0 [coresight] [ 423.616589] coresight_unregister+0x120/0x170 [coresight] [ 423.623533] arm_trbe_remove_coresight_cpu+0x70/0xa0 [coresight_trbe] [ 423.631082] __flush_smp_call_function_queue+0x1e4/0x4e0 [ 423.637471] generic_smp_call_function_single_interrupt+0x1c/0x30 [ 423.644796] ipi_handler+0x90/0x278 [ 423.648992] handle_percpu_devid_irq+0x90/0x250 [ 423.654636] generic_handle_domain_irq+0x34/0x58 [ 423.659786] gic_handle_irq+0x12c/0x270 [ 423.664039] call_on_irq_stack+0x24/0x30 [ 423.668452] do_interrupt_handler+0x88/0x98 [ 423.673027] el1_interrupt+0x48/0xe8 [ 423.677413] el1h_64_irq_handler+0x18/0x28 [ 423.681781] el1h_64_irq+0x78/0x80 [ 423.685550] default_idle_call+0x5c/0x180 [ 423.689855] do_idle+0x25c/0x2c0 [ 423.694196] cpu_startup_entry+0x2c/0x40 [ 423.698373] secondary_start_kernel+0x144/0x188 [ 423.703920] __secondary_switched+0xb8/0xc0 [ 423.708972] ---[ end trace 0000000000000000 ]--- [ 423.729209] ------------[ cut here ]------------ ... [ 423.735217] WARNING: CPU: 2 PID: 40 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ... [ 424.012385] WARNING: CPU: 3 PID: 0 at drivers/base/devres.c:1064 devm_kfree+0x88/0x98 ...Quoted 7 lines
Quoted 5 lines
This patch does the following:
1.TRBE coresight devices do not need regular connections information, We can free connections resource when thenr_connsis valid. 2.And we can ignore the free platform data resource, it will be automatically free inplatform_driver_unregister().Do we need a Fixes tag here ?
Yes, I will do that.
Quoted 31 lines
Quoted 12 lines
Signed-off-by Junhao He <[email protected]>---
drivers/hwtracing/coresight/coresight-core.c| 7 ++++---1 file changed, 4 insertions(+), 3 deletions(-)drivers/hwtracing/coresight/coresight-core.c +0
-0@@ -1555,9 +1555,10 @@ void coresight_release_platform_data(struct coresight_device *csdev,drivers/hwtracing/coresight/coresight-core.c void coresight_release_platform_data(struct · line 1562
Quoted 10 lines
drivers/hwtracing/coresight/coresight-core.c +4
-3conns[i]->dest_fwnode = NULL; devm_kfree(dev, conns[i]); } - devm_kfree(dev, pdata->out_conns); - devm_kfree(dev, pdata->in_conns); - devm_kfree(dev, pdata); + if (pdata->nr_outconns) + devm_kfree(dev, pdata->out_conns); + if (pdata->nr_inconns) + devm_kfree(dev, pdata->in_conns);These allocations are made on the parent device and that may never get unregistered (e.g., AMBA device, platform device, stay forever, even when the "coresight" modules are unloaded). Thus the memory will be left unused, literally leaking. This specific
devm_kfree()was added to fix that. May be we should fix this in the TRBE driver to use separate pdata for the TRBE device instances.Quoted 1 line
Suzuki
Quoted 7 lines
If we fix this with minimal changes, I think it is possible to add a check and not free pdata if it is TRBE?
if (
csdev->subtype.sink_subtype!=CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM)devm_kfree(dev, pdata);Then free pdata in the end of
arm_trbe_remove_coresight().It is much nicer to do something like :
--8>--
coresight: trbe: Allocate platform data per device
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit
4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata percoresight_device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Reported-by Junhao He <[email protected]>Cc Anshuman Khandual <[email protected]>Signed-off-by Suzuki K Poulose <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +0
-0@@ -1255,10 +1255,17 @@ static void
Quoted 3 lines
arm_trbe_register_coresight_cpu(struct
trbe_drvdata*drvdata, int cp
if (!desc.name)
gotocpu_clear;
---8>--
Quoted 5 lines
+ /*
+ * TRBE doesn't have any connections described via firmware. Instead + * we register the trbe instance as per-cpu sink.
+ */
Ignore the comments above ^. I have tested this locally and works for me. Please could you confirm if this solves the problem for you ?
Kind regards
Suzuki
Quoted 33 lines
+ desc.pdata =
coresight_get_platform_data(dev); + if (IS_ERR(desc.pdata))
+ gotocpu_clear;drivers/hwtracing/coresight/coresight-trbe.c +1
-2+ desc.type = CORESIGHT_DEV_TYPE_SINK; desc.subtype.sink_subtype = CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM; desc.ops = &arm_trbe_cs_ops; - desc.pdata = dev_get_platdata(dev); desc.groups = arm_trbe_groups; desc.dev = dev; trbe_csdev = coresight_register(&desc);@@ -1482,7 +1489,6 @@ static void arm_trbe_remove_irq(struct trbe_drvdata *drvdata) static int arm_trbe_device_probe(struct platform_device *pdev) { - struct coresight_platform_data *pdata; struct trbe_drvdata *drvdata; struct device *dev = &pdev->dev;int ret;
drivers/hwtracing/coresight/coresight-trbe.c +0
-5@@ -1497,12 +1503,7 @@ static int arm_trbe_device_probe(struct platform_device *pdev) if (!drvdata) return -ENOMEM; - pdata = coresight_get_platform_data(dev); - if (IS_ERR(pdata)) - return PTR_ERR(pdata); - dev_set_drvdata(dev, drvdata); - dev->platform_data = pdata; drvdata->pdev = pdev; ret = arm_trbe_probe_irq(pdev, drvdata);
Quoted 1 line
if (ret)
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerPATCH v2 1/2 coresight: trbe: Fix TRBE potential sleep in atomic context
In reply to Junhao HeFrom: Junhao He <[email protected]>
smp_call_function_single() will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will execute arm_trbe_remove_coresight_cpu() call request in the interrupt handler.
According to the device_unregister() stack information, if other process is useing the device, the down_write() may sleep, and trigger deadlocks or unexpected errors.
arm_trbe_remove_coresight_cpu
coresight_unregister
device_unregister
device_del
kobject_del
__kobject_del
sysfs_remove_dir
kernfs_remove
down_write ---------> it may sleepAdd a helper arm_trbe_disable_cpu() to disable TRBE precpu irq and reset per TRBE.
Simply call arm_trbe_remove_coresight_cpu() directly without useing the smp_call_function_single(), which is the same as registering the TRBE coresight device.
--- drivers/hwtracing/coresight/coresight-trbe.c | 33 +++++++++++--------- 1 file changed, 18 insertions(+), 15 deletions(-)
drivers/hwtracing/coresight/coresight-trbe.c +18 -15
@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info)enable_percpu_irq (drvdata->irq,IRQ_TYPE_NONE ); } +static void arm_trbe_disable_cpu (void *info) +{ +struct trbe_drvdata *drvdata = info; +struct trbe_cpudata *cpudata =this_cpu_ptr (drvdata->cpudata); + +disable_percpu_irq (drvdata->irq); +trbe_reset_local (cpudata); + cpudata->drvdata =NULL ; +} + +static void arm_trbe_register_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) {struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);
@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info)cpumask_clear_cpu (cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu (void *info) +static void arm_trbe_remove_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) { -int cpu =smp_processor_id (); -struct trbe_drvdata *drvdata = info; -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);struct coresight_device *trbe_csdev =coresight_get_percpu_sink (cpu); -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata);if (trbe_csdev) {coresight_unregister (trbe_csdev); - cpudata->drvdata =NULL ;coresight_set_percpu_sink (cpu,NULL ); } }
@@ -1366,8 +1371,10 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) {int cpu; -for_each_cpu (cpu, &drvdata->supported_cpus) -smp_call_function_single (cpu, arm_trbe_remove_coresight_cpu, drvdata,1 ); +for_each_cpu (cpu, &drvdata->supported_cpus) { +smp_call_function_single (cpu, arm_trbe_disable_cpu, drvdata,1 ); +arm_trbe_remove_coresight_cpu (drvdata, cpu); + }free_percpu (drvdata->cpudata);return 0 ; }
@@ -1406,12 +1413,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) { struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node); -if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) { -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu); - -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata); - } +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu (drvdata);return 0 ; }
Signature · 7 lines
2.34.1 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerPATCH 2/2 coresight: trbe: Allocate platform data per device
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit 4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata per coresight_device.
--- drivers/hwtracing/coresight/coresight-trbe.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)
drivers/hwtracing/coresight/coresight-trbe.c +4 -7
@@ -1255,10 +1255,13 @@ static void arm_trbe_register_coresight_cpu(struct trbe_drvdata *drvdata, int cpif (!desc.name) goto cpu_clear; + desc.pdata =coresight_get_platform_data (dev); +if (IS_ERR (desc.pdata)) + goto cpu_clear; + desc.type =CORESIGHT_DEV_TYPE_SINK ; desc.subtype.sink_subtype =CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM ; desc.ops = &arm_trbe_cs_ops; - desc.pdata =dev_get_platdata (dev); desc.groups = arm_trbe_groups; desc.dev = dev; trbe_csdev =coresight_register (&desc);
@@ -1482,7 +1485,6 @@ static void arm_trbe_remove_irq(struct trbe_drvdata *drvdata)static int arm_trbe_device_probe (struct platform_device *pdev) { -struct coresight_platform_data *pdata;struct trbe_drvdata *drvdata;struct device *dev = &pdev->dev;int ret;
@@ -1497,12 +1499,7 @@ static int arm_trbe_device_probe(struct platform_device *pdev)if (!drvdata)return -ENOMEM ; - pdata =coresight_get_platform_data (dev); -if (IS_ERR (pdata)) -return PTR_ERR (pdata); -dev_set_drvdata (dev, drvdata); - dev->platform_data = pdata; drvdata->pdev = pdev; ret =arm_trbe_probe_irq (pdev, drvdata);if (ret)
Signature · 1 line
2.34.1
Re: [PATCH 2/2] coresight: trbe: Allocate platform data per device
Hi Suzuki,
Seems like this patch is going to conflict with the below proposed change
https://lore.kernel.org/all/[email protected]/
Please let me know how should we resolve this conflict.
Quoted 8 lines · Suzuki K Poulose
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit
4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata percoresight_device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")
The above mentioned commit i.e 4e8fe7e5c3a5 seems to be a more recent one which has triggered this problem. But would the problem be still there without that ? Else 'Fixes:' tag would need changing.
Quoted 38 lines
Reported-by Junhao He <[email protected]>Cc Anshuman Khandual <[email protected]>Signed-off-by Suzuki K Poulose <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +4
-2@@ -1255,10 +1255,13 @@ static void arm_trbe_register_coresight_cpu(struct trbe_drvdata *drvdata, int cpif (!desc.name) goto cpu_clear; + desc.pdata =coresight_get_platform_data (dev); +if (IS_ERR (desc.pdata)) + goto cpu_clear; + desc.type =CORESIGHT_DEV_TYPE_SINK ; desc.subtype.sink_subtype =CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM ; desc.ops = &arm_trbe_cs_ops; - desc.pdata =dev_get_platdata (dev); desc.groups = arm_trbe_groups; desc.dev = dev; trbe_csdev =coresight_register (&desc);@@ -1482,7 +1485,6 @@ static void arm_trbe_remove_irq(struct trbe_drvdata *drvdata)static int arm_trbe_device_probe (struct platform_device *pdev) { -struct coresight_platform_data *pdata;struct trbe_drvdata *drvdata;struct device *dev = &pdev->dev;int ret;@@ -1497,12 +1499,7 @@ static int arm_trbe_device_probe(struct platform_device *pdev)if (!drvdata)return -ENOMEM ;
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_device_probe(struct platform_device *pdev) · line 1505
Quoted 10 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-5- pdata =coresight_get_platform_data (dev); -if (IS_ERR (pdata)) -return PTR_ERR (pdata); -dev_set_drvdata (dev, drvdata); - dev->platform_data = pdata; drvdata->pdev = pdev; ret =arm_trbe_probe_irq (pdev, drvdata);if (ret)
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 2/2] coresight: trbe: Allocate platform data per device
Quoted 7 lines · Anshuman Khandual
Hi Suzuki,
Seems like this patch is going to conflict with the below proposed change
https://lore.kernel.org/all/[email protected]/
Please let me know how should we resolve this conflict.
We could merge them both, with the fixes: one first, just to acknowledge that there was a problem. But I suppose your one will have to be rebased on top.
Quoted 9 lines · Suzuki K Poulose
Quoted 7 lines
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit
4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata percoresight_device.
Quoted 6 lines
Quoted 1 line
Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")The above mentioned commit i.e
4e8fe7e5c3a5seems to be a more recent one which has triggered this problem. But would the problem be still there without that ? Else 'Fixes:' tag would need changing.
Yes I think the fixes tag should point to 4e8fe7e5c3a5.
Quoted 39 lines
Quoted 39 lines
Reported-by Junhao He <[email protected]>Cc Anshuman Khandual <[email protected]>Signed-off-by Suzuki K Poulose <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +4
-2@@ -1255,10 +1255,13 @@ static void arm_trbe_register_coresight_cpu(struct trbe_drvdata *drvdata, int cpif (!desc.name) goto cpu_clear; + desc.pdata =coresight_get_platform_data (dev); +if (IS_ERR (desc.pdata)) + goto cpu_clear; + desc.type =CORESIGHT_DEV_TYPE_SINK ; desc.subtype.sink_subtype =CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM ; desc.ops = &arm_trbe_cs_ops; - desc.pdata =dev_get_platdata (dev); desc.groups = arm_trbe_groups; desc.dev = dev; trbe_csdev =coresight_register (&desc);@@ -1482,7 +1485,6 @@ static void arm_trbe_remove_irq(struct trbe_drvdata *drvdata)static int arm_trbe_device_probe (struct platform_device *pdev) { -struct coresight_platform_data *pdata;struct trbe_drvdata *drvdata;struct device *dev = &pdev->dev;int ret;@@ -1497,12 +1499,7 @@ static int arm_trbe_device_probe(struct platform_device *pdev)if (!drvdata)return -ENOMEM ;
Quoted 9 lines
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_device_probe(struct platform_device *pdev) · line 1505
Quoted 9 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-5- pdata =coresight_get_platform_data (dev); -if (IS_ERR (pdata)) -return PTR_ERR (pdata); -dev_set_drvdata (dev, drvdata); - dev->platform_data = pdata; drvdata->pdev = pdev; ret =arm_trbe_probe_irq (pdev, drvdata);if (ret)
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 2/2] coresight: trbe: Allocate platform data per device
Quoted 24 lines · James Clark
Quoted 7 lines · Anshuman Khandual
Hi Suzuki,
Seems like this patch is going to conflict with the below proposed change
https://lore.kernel.org/all/[email protected]/
Please let me know how should we resolve this conflict.
We could merge them both, with the fixes: one first, just to acknowledge that there was a problem. But I suppose your one will have to be rebased on top.
Quoted 9 lines · Suzuki K Poulose
Quoted 7 lines
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit
4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata percoresight_device.
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")The above mentioned commit i.e
4e8fe7e5c3a5seems to be a more recent one which has triggered this problem. But would the problem be still there without that ? Else 'Fixes:' tag would need changing.Yes I think the fixes tag should point to
4e8fe7e5c3a5.
Agreed, I will change the fixes tag and push this.
Suzuki
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 2/2] coresight: trbe: Allocate platform data per device
Quoted 27 lines · Suzuki K Poulose
Quoted 26 lines · James Clark
Quoted 7 lines · Anshuman Khandual
Hi Suzuki,
Seems like this patch is going to conflict with the below proposed change
https://lore.kernel.org/all/[email protected]/
Please let me know how should we resolve this conflict.
We could merge them both, with the fixes: one first, just to acknowledge that there was a problem. But I suppose your one will have to be rebased on top.
Quoted 11 lines · Suzuki K Poulose
Quoted 8 lines
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit
4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata percoresight_device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
The above mentioned commit i.e
4e8fe7e5c3a5seems to be a more recent one which has triggered this problem. But would the problem be still there without that ? Else 'Fixes:' tag would need changing.Yes I think the fixes tag should point to
4e8fe7e5c3a5.Agreed, I will change the fixes tag and push this.
In the first patch, the last hunk might not be required to fix the IPI problem and in fact might be bit problematic as well. Besides, could you please hold off pushing this change into coresight tree for some time ?
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 2/2] coresight: trbe: Allocate platform data per device
Quoted 35 lines · Anshuman Khandual
Quoted 32 lines · Suzuki K Poulose
Quoted 31 lines · James Clark
Quoted 7 lines · Anshuman Khandual
Hi Suzuki,
Seems like this patch is going to conflict with the below proposed change
https://lore.kernel.org/all/[email protected]/
Please let me know how should we resolve this conflict.
We could merge them both, with the fixes: one first, just to acknowledge that there was a problem. But I suppose your one will have to be rebased on top.
Quoted 15 lines · Suzuki K Poulose
Quoted 8 lines
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit
4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata percoresight_device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")The above mentioned commit i.e
4e8fe7e5c3a5seems to be a more recent one which has triggered this problem. But would the problem be still there without that ? Else 'Fixes:' tag would need changing.
Quoted 6 lines
Quoted 3 lines
Quoted 1 line
Yes I think the fixes tag should point to
4e8fe7e5c3a5.Agreed, I will change the fixes tag and push this.
In the first patch, the last hunk might not be required to fix the IPI problem and in fact might be bit problematic as well. Besides,
Please could you comment your concerns on the patch ?
Suzuki
Quoted 2 lines
could you please hold off pushing this change into coresight tree for some time ?
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH 2/2] coresight: trbe: Allocate platform data per device
In reply to Anshuman KhandualQuoted 7 lines · Anshuman Khandual
Hi Suzuki,
Seems like this patch is going to conflict with the below proposed change
https://lore.kernel.org/all/[email protected]/
Please let me know how should we resolve this conflict.
Please rebase your change on top of this one.
Suzuki
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 2/2] coresight: trbe: Allocate platform data per device
In reply to Suzuki K PouloseQuoted 7 lines · Suzuki K Poulose
Coresight TRBE driver shares a single platform data (which is empty btw). However, with the commit
4e8fe7e5c3a5
("coresight: Store pointers to connections rather than an array of them") the coresight core would free up the pdata, resulting in multiple attempts to free the same pdata for TRBE instances. Fix this by allocating a pdata percoresight_device.
Quoted 5 lines
Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Reported-by Junhao He <[email protected]>Cc Anshuman Khandual <[email protected]>Signed-off-by Suzuki K Poulose <[email protected]>
Quoted 34 lines
--- drivers/hwtracing/coresight/coresight-trbe.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +4
-2@@ -1255,10 +1255,13 @@ static void arm_trbe_register_coresight_cpu(struct trbe_drvdata *drvdata, int cpif (!desc.name) goto cpu_clear; + desc.pdata =coresight_get_platform_data (dev); +if (IS_ERR (desc.pdata)) + goto cpu_clear; + desc.type =CORESIGHT_DEV_TYPE_SINK ; desc.subtype.sink_subtype =CORESIGHT_DEV_SUBTYPE_SINK_PERCPU_SYSMEM ; desc.ops = &arm_trbe_cs_ops; - desc.pdata =dev_get_platdata (dev); desc.groups = arm_trbe_groups; desc.dev = dev; trbe_csdev =coresight_register (&desc);@@ -1482,7 +1485,6 @@ static void arm_trbe_remove_irq(struct trbe_drvdata *drvdata)static int arm_trbe_device_probe (struct platform_device *pdev) { -struct coresight_platform_data *pdata;struct trbe_drvdata *drvdata;struct device *dev = &pdev->dev;int ret;@@ -1497,12 +1499,7 @@ static int arm_trbe_device_probe(struct platform_device *pdev)if (!drvdata)return -ENOMEM ;
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_device_probe(struct platform_device *pdev) · line 1505
Quoted 10 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-5- pdata =coresight_get_platform_data (dev); -if (IS_ERR (pdata)) -return PTR_ERR (pdata); -dev_set_drvdata (dev, drvdata); - dev->platform_data = pdata; drvdata->pdev = pdev; ret =arm_trbe_probe_irq (pdev, drvdata);if (ret)
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH v2 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
In reply to Suzuki K PouloseHello Junhao,
Quoted 13 lines · Suzuki K Poulose
From: Junhao He <[email protected]>
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.
arm_trbe_remove_coresight_cpu
Quoted 8 lines
coresight_unregister
device_unregister
device_del
kobject_del
__kobject_del
sysfs_remove_dir
kernfs_remove
down_write ---------> it may sleepBut how did you really detect this problem ? Does this show up as an warning when you enable lockdep debug ? OR it really happened during a real workload execution followed by TRBE module unload. Although the problem seems plausible (which needs fixing), just wondering how did we trigger this.
Quoted 73 lines
Add a helper
arm_trbe_disable_cpu()to disable TRBE precpu irq and reset per TRBE.
Simply callarm_trbe_remove_coresight_cpu()directly without useing thesmp_call_function_single(), which is the same as registering the TRBE coresight device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Signed-off-by Junhao He <[email protected]>[ Remove duplicate cpumask checks during removal ]Signed-off-by Suzuki K Poulose <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 33 +++++++++++--------- 1 file changed, 18 insertions(+), 15 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +16
-9@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info)enable_percpu_irq (drvdata->irq,IRQ_TYPE_NONE ); } +static void arm_trbe_disable_cpu (void *info) +{ +struct trbe_drvdata *drvdata = info; +struct trbe_cpudata *cpudata =this_cpu_ptr (drvdata->cpudata); + +disable_percpu_irq (drvdata->irq); +trbe_reset_local (cpudata); + cpudata->drvdata =NULL ; +} + +static void arm_trbe_register_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) {struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info)cpumask_clear_cpu (cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu (void *info) +static void arm_trbe_remove_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) { -int cpu =smp_processor_id (); -struct trbe_drvdata *drvdata = info; -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);struct coresight_device *trbe_csdev =coresight_get_percpu_sink (cpu); -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata);if (trbe_csdev) {coresight_unregister (trbe_csdev); - cpudata->drvdata =NULL ;coresight_set_percpu_sink (cpu,NULL ); } }@@ -1366,8 +1371,10 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) {int cpu; -for_each_cpu (cpu, &drvdata->supported_cpus) -smp_call_function_single (cpu, arm_trbe_remove_coresight_cpu, drvdata,1 ); +for_each_cpu (cpu, &drvdata->supported_cpus) { +smp_call_function_single (cpu, arm_trbe_disable_cpu, drvdata,1 ); +arm_trbe_remove_coresight_cpu (drvdata, cpu); + }free_percpu (drvdata->cpudata);return 0 ; }@@ -1406,12 +1413,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) {
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) · line 1417
Quoted 10 lines
drivers/hwtracing/coresight/coresight-trbe.c +2
-6struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node); -if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) { -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu); - -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata); - } +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu (drvdata);
This code hunk seems unrelated to the context here other than just finding another use case for arm_trbe_disable_cpu(). The problem is - arm_trbe_disable_cpu() resets cpudata->drvdata which might not get re-initialized back in arm_trbe_cpu_startup(), as there will still be a per cpu sink associated as confirmed with coresight_get_percpu_sink(). I guess it might be better to drop this change and just keep everything limited to SMP IPI callback reworking in arm_trbe_remove_coresight().
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) · line 1420
Quoted 3 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-0return 0 ; }
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH v2 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
Hi Anshuman Khandual,
Quoted 24 lines · Anshuman Khandual
Hello Junhao,
Quoted 12 lines · Suzuki K Poulose
From: Junhao He <[email protected]>
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.Quoted 9 lines
arm_trbe_remove_coresight_cpu coresight_unregister device_unregister device_del kobject_del __kobject_del sysfs_remove_dir kernfs_remove down_write ---------> it may sleep
Quoted 4 lines
But how did you really detect this problem ? Does this show up as an warning when you enable lockdep debug ? OR it really happened during a real workload execution followed by TRBE module unload. Although the problem seems plausible (which needs fixing), just wondering how did we trigger this.
Yes, it really happened during a real workload.
If the TRBE driver is loaded and unloaded cyclically. the test script following:
for ((i=0;i<99999;i++))
do
insmod coresight-trbe.ko;
rmmod coresight-trbe.ko;
echo "loop $i";
done
The kernel will report a panic.
Quoted 74 lines
Quoted 74 lines
Add a helper
arm_trbe_disable_cpu()to disable TRBE precpu irq and reset per TRBE.
Simply callarm_trbe_remove_coresight_cpu()directly without useing thesmp_call_function_single(), which is the same as registering the TRBE coresight device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Signed-off-by Junhao He <[email protected]>[ Remove duplicate cpumask checks during removal ]Signed-off-by Suzuki K Poulose <[email protected]>--- drivers/hwtracing/coresight/coresight-trbe.c | 33 +++++++++++--------- 1 file changed, 18 insertions(+), 15 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +16
-9@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info)enable_percpu_irq (drvdata->irq,IRQ_TYPE_NONE ); } +static void arm_trbe_disable_cpu (void *info) +{ +struct trbe_drvdata *drvdata = info; +struct trbe_cpudata *cpudata =this_cpu_ptr (drvdata->cpudata); + +disable_percpu_irq (drvdata->irq); +trbe_reset_local (cpudata); + cpudata->drvdata =NULL ; +} + +static void arm_trbe_register_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) {struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info)cpumask_clear_cpu (cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu (void *info) +static void arm_trbe_remove_coresight_cpu (struct trbe_drvdata *drvdata,int cpu) { -int cpu =smp_processor_id (); -struct trbe_drvdata *drvdata = info; -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu);struct coresight_device *trbe_csdev =coresight_get_percpu_sink (cpu); -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata);if (trbe_csdev) {coresight_unregister (trbe_csdev); - cpudata->drvdata =NULL ;coresight_set_percpu_sink (cpu,NULL ); } }@@ -1366,8 +1371,10 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) {int cpu; -for_each_cpu (cpu, &drvdata->supported_cpus) -smp_call_function_single (cpu, arm_trbe_remove_coresight_cpu, drvdata,1 ); +for_each_cpu (cpu, &drvdata->supported_cpus) { +smp_call_function_single (cpu, arm_trbe_disable_cpu, drvdata,1 ); +arm_trbe_remove_coresight_cpu (drvdata, cpu); + }free_percpu (drvdata->cpudata);return 0 ; }@@ -1406,12 +1413,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) { struct trbe_drvdata *drvdata = hlist_entry_safe(node, struct trbe_drvdata, hotplug_node);
Quoted 14 lines
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) · line 1417
Quoted 8 lines
drivers/hwtracing/coresight/coresight-trbe.c +2
-6-if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) { -struct trbe_cpudata *cpudata =per_cpu_ptr (drvdata->cpudata, cpu); - -disable_percpu_irq (drvdata->irq); -trbe_reset_local (cpudata); - } +if (cpumask_test_cpu (cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu (drvdata);This code hunk seems unrelated to the context here other than just finding another use case for
arm_trbe_disable_cpu(). The problem is -arm_trbe_disable_cpu()resetscpudata->drvdatawhich might not get re-initialized back inarm_trbe_cpu_startup(), as there will still be a per cpu sink associated as confirmed withcoresight_get_percpu_sink(). I guess it might be better to drop this change and just keep everything limited to SMP IPI callback reworking inarm_trbe_remove_coresight().
OK, will fix it. The change is just to simplify the code of cpu_teardown. Maybe we can consider whether we need to set "cpudata->drvdata = NULL" in arm_trbe_disable_cpu()? If it's not necessary, This can be kept. Then drop the release cpudata->drvdata from arm_trbe_disable_cpu().
Best regards,
Junhao.
Quoted 5 lines
drivers/hwtracing/coresight/coresight-trbe.c static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) · line 1420
Quoted 3 lines
drivers/hwtracing/coresight/coresight-trbe.c +0
-0return 0 ; }.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH v2 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
Quoted 41 lines · hejunhao
Hi Anshuman Khandual,
Quoted 24 lines · Anshuman Khandual
Hello Junhao,
Quoted 12 lines · Suzuki K Poulose
From: Junhao He <[email protected]>
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.Quoted 9 lines
arm_trbe_remove_coresight_cpucoresight_unregister
device_unregister
device_del
kobject_del
__kobject_del
sysfs_remove_dir
kernfs_remove
down_write---------> it may sleepQuoted 7 lines
But how did you really detect this problem ? Does this show up as an warning when
you enable lockdep debug ? OR it really happened during a real workload execution
followed by TRBE module unload. Although the problem seems plausible (which needs
fixing), just wondering how did we trigger this.Yes, it really happened during a real workload.
If the TRBE driver is loaded and unloaded cyclically. the test script following:
Quoted 9 lines
for ((i=0;i<99999;i++))
do
insmod coresight-trbe.ko;
rmmod coresight-trbe.ko;
echo "loop $i";
doneThe kernel will report a panic.
I wonder how easy it would be to add a kselftest to do this with all of the Coresight modules. Because we also had a problem with bad reference counting preventing an unload of the CTI module. Although that did require starting a perf session, which might complicated the test.
Quoted 111 lines
Quoted 91 lines
Quoted 80 lines
Add a helper
arm_trbe_disable_cpu()to disable TRBE precpu irq and reset per TRBE.
Simply callarm_trbe_remove_coresight_cpu()directly without useing thesmp_call_function_single(), which is the same as registering the TRBE coresight device.Fixes 3fbf7f011f24 ("coresight: sink: Add TRBE driver")Signed-off-by Junhao He <[email protected]>Link:
https://lore.kernel.org/r/[email protected][ Remove duplicate cpumask checks during removal ]Signed-off-by Suzuki K Poulose <[email protected]>---
drivers/hwtracing/coresight/coresight-trbe.c| 33 +++++++++++---------1 file changed, 18 insertions(+), 15 deletions(-)drivers/hwtracing/coresight/coresight-trbe.c +10
-0@@ -1225,6 +1225,17 @@ static void arm_trbe_enable_cpu(void *info) enable_percpu_irq(drvdata->irq, IRQ_TYPE_NONE); } +static void arm_trbe_disable_cpu(void *info) +{ + struct trbe_drvdata *drvdata = info; + struct trbe_cpudata *cpudata = this_cpu_ptr(drvdata->cpudata); + + disable_percpu_irq(drvdata->irq); + trbe_reset_local(cpudata); + cpudata->drvdata = NULL; +} + +static void
arm_trbe_register_coresight_cpu(struct trbe_drvdata *drvdata, int cpu)
{
structtrbe_cpudata*cpudata =per_cpu_ptr(drvdata->cpudata, cpu);drivers/hwtracing/coresight/coresight-trbe.c +1
-3@@ -1326,18 +1337,12 @@ static void arm_trbe_probe_cpu(void *info) cpumask_clear_cpu(cpu, &drvdata->supported_cpus); } -static void arm_trbe_remove_coresight_cpu(void *info) +static void arm_trbe_remove_coresight_cpu(struct trbe_drvdata *drvdata, int cpu) { - int cpu = smp_processor_id(); - struct trbe_drvdata *drvdata = info; - struct trbe_cpudata *cpudata = per_cpu_ptr(drvdata->cpudata, cpu);struct
coresight_device*trbe_csdev=coresight_get_percpu_sink(cpu);
-disable_percpu_irq(drvdata->irq);
-trbe_reset_local(cpudata);
if (trbe_csdev) {
coresight_unregister(trbe_csdev);
-cpudata->drvdata= NULL;
coresight_set_percpu_sink(cpu, NULL);
}
}drivers/hwtracing/coresight/coresight-trbe.c +4
-1@@ -1366,8 +1371,10 @@ static int arm_trbe_remove_coresight(struct trbe_drvdata *drvdata) { int cpu; - for_each_cpu(cpu, &drvdata->supported_cpus) - smp_call_function_single(cpu, arm_trbe_remove_coresight_cpu, drvdata, 1); + for_each_cpu(cpu, &drvdata->supported_cpus) { + smp_call_function_single(cpu, arm_trbe_disable_cpu, drvdata,1 ); + arm_trbe_remove_coresight_cpu(drvdata, cpu); + }
free_percpu(drvdata->cpudata);
return 0;
}drivers/hwtracing/coresight/coresight-trbe.c +0
-0@@ -1406,12 +1413,8 @@ static int arm_trbe_cpu_teardown(unsigned int cpu, struct hlist_node *node) {struct
trbe_drvdata*drvdata =hlist_entry_safe(node, struct trbe_drvdata, hotplug_node);
- if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) {Quoted 8 lines
- struct
trbe_cpudata*cpudata =per_cpu_ptr(drvdata->cpudata, cpu);
-
-disable_percpu_irq(drvdata->irq);
-trbe_reset_local(cpudata);
- }
+ if (cpumask_test_cpu(cpu, &drvdata->supported_cpus)) +arm_trbe_disable_cpu(drvdata);This code hunk seems unrelated to the context here other than just finding another use case
forarm_trbe_disable_cpu(). The problem is -arm_trbe_disable_cpu()Quoted 8 lines
resets
cpudata->drvdata
which might not get re-initialized back inarm_trbe_cpu_startup(), as there will still be a
per cpu sink associated as confirmed withcoresight_get_percpu_sink(). I guess it might be
better to drop this change and just keep everything limited to SMP IPI callback reworking inarm_trbe_remove_coresight().OK, will fix it. The change is just to simplify the code of
cpu_teardown. Maybe we can consider whether we need to set "cpudata->drvdata= NULL" in arm_trbe_disable_cpu()? If it's not necessary, This can be kept. Then drop the releasecpudata->drvdatafromarm_trbe_disable_cpu().Best regards,
Junhao.Quoted 3 lines
Quoted 3 lines
return 0;
}
Quoted 6 lines
Quoted 2 lines
.
_______________________________________________
CoreSight mailing list -- [email protected] To unsubscribe send an email to [email protected]
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Suzuki K Poulose
maintainerRe: [PATCH v2 1/2] coresight: trbe: Fix TRBE potential sleep in atomic context
In reply to Anshuman KhandualQuoted 25 lines · Anshuman Khandual
Hello Junhao,
Quoted 12 lines · Suzuki K Poulose
From: Junhao He <[email protected]>
smp_call_function_single()will allocate an IPI interrupt vector to the target processor and send a function call request to the interrupt vector. After the target processor receives the IPI interrupt, it will executearm_trbe_remove_coresight_cpu()call request in the interrupt handler.According to the
device_unregister()stack information, if other process is useing the device, thedown_write()may sleep, and trigger deadlocks or unexpected errors.Quoted 9 lines
arm_trbe_remove_coresight_cpu coresight_unregister device_unregister device_del kobject_del __kobject_del sysfs_remove_dir kernfs_remove down_write ---------> it may sleep
Quoted 4 lines
But how did you really detect this problem ? Does this show up as an warning when you enable lockdep debug ? OR it really happened during a real workload execution followed by TRBE module unload. Although the problem seems plausible (which needs fixing), just wondering how did we trigger this.
I was able to trigger this with just :
modprobe coresight-trbe; modprobe -r coresight-trbe;
With all the bells and whistles enabled in the kernel.
Suzuki
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Quoted 8 lines · Junhao He
The TRBE driver support is build as a module, we found some driver issues based on the patchset [1] and set
CONFIG_CORESIGHT_TRBE=m.
1. TRBE driver potential sleep in atomic context when unregister device
2. Multiple free the platform data resource when rmmod coresight TRBE driver[1] "coresight: trbe: Enable ACPI based devices"
https://lore.kernel.org/all/[email protected]/
I am glad that ACPI based device enablement is already helping in getting more test coverage for the TRBE driver itself. I have just posted a new version for the above series.
https://lore.kernel.org/all/[email protected]/
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel