Skip to contentSkip to conversation

device_cgroup: Fix RCU list debugging warning

PATCH In review Sasha Levin 2 participants 17 messages Sep 21, 2020 – Sep 22, 2020

Subsystems (6)
lore
Share & tools
lore

Sasha Levin

author maintainer
#0 lore raw

PATCH 1/15 device_cgroup: Fix RCU list debugging warning

From: Amol Grover <[email protected]>

[ Upstream commit bc62d68e2a0a69fcdcf28aca8edb01abf306b698 ]

exceptions may be traversed using list_for_each_entry_rcu() outside of an RCU read side critical section BUT under the protection of decgroup_mutex. Hence add the corresponding lockdep expression to fix the following false-positive warning:

[    2.304417] =============================
[    2.304418] WARNING: suspicious RCU usage
[    2.304420] 5.5.4-stable #17 Tainted: G            E
[    2.304422] -----------------------------
[    2.304424] security/device_cgroup.c:355 RCU-list traversed in non-reader section!!
---
 security/device_cgroup.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
security/device_cgroup.c +2 -1
index 725674f3276d3..5d7bb91c64876 100644
@@ -352,7 +352,8 @@ static bool match_exception_partial(struct list_head *exceptions, short type,
 {
 	struct dev_exception_item *ex;
 
-	list_for_each_entry_rcu(ex, exceptions, list) {
+	list_for_each_entry_rcu(ex, exceptions, list,
+				lockdep_is_held(&devcgroup_mutex)) {
 		if ((type & DEVCG_DEV_BLOCK) && !(ex->type & DEVCG_DEV_BLOCK))
 			continue;
 		if ((type & DEVCG_DEV_CHAR) && !(ex->type & DEVCG_DEV_CHAR))
Signature · 1 line
2.25.1
author maintainer
#1 lore raw

PATCH 2/15 ASoC: pcm3168a: ignore 0 Hz settings

From: Kuninori Morimoto <[email protected]>

[ Upstream commit 7ad26d6671db758c959d7e1d100b138a38483612 ]

Some sound card try to set 0 Hz as reset, but it is impossible. This patch ignores it to avoid error return.

---
 sound/soc/codecs/pcm3168a.c | 7 +++++++
 1 file changed, 7 insertions(+)
sound/soc/codecs/pcm3168a.c +7 -0
index 88b75695fbf7f..b37e5fbbd301a 100644
@@ -302,6 +302,13 @@ static int pcm3168a_set_dai_sysclk(struct snd_soc_dai *dai,
 	struct pcm3168a_priv *pcm3168a = snd_soc_component_get_drvdata(dai->component);
 	int ret;
 
+	/*
+	 * Some sound card sets 0 Hz as reset,
+	 * but it is impossible to set. Ignore it here
+	 */
+	if (freq == 0)
+		return 0;
+
 	if (freq > PCM3168A_MAX_SYSCLK)
 		return -EINVAL;
 
Signature · 1 line
2.25.1
author maintainer
#2 lore raw

PATCH 3/15 ASoC: wm8994: Skip setting of the WM8994_MICBIAS register for WM1811

In reply to Sasha Levin

From: Sylwester Nawrocki <[email protected]>

[ Upstream commit 811c5494436789e7149487c06e0602b507ce274b ]

The WM8994_MICBIAS register is not available in the WM1811 CODEC so skip initialization of that register for that device.
This suppresses an error during boot:
"wm8994-codec: ASoC: error at snd_soc_component_update_bits on wm8994-codec"

---
 sound/soc/codecs/wm8994.c  | 2 ++
 sound/soc/codecs/wm_hubs.c | 3 +++
 sound/soc/codecs/wm_hubs.h | 1 +
 3 files changed, 6 insertions(+)
sound/soc/codecs/wm8994.c +2 -0
index d5fb7f5dd551c..64635f9cdae65 100644
@@ -4047,11 +4047,13 @@ static int wm8994_component_probe(struct snd_soc_component *component)
 			wm8994->hubs.dcs_readback_mode = 2;
 			break;
 		}
+		wm8994->hubs.micd_scthr = true;
 		break;
 
 	case WM8958:
 		wm8994->hubs.dcs_readback_mode = 1;
 		wm8994->hubs.hp_startup_mode = 1;
+		wm8994->hubs.micd_scthr = true;
 
 		switch (control->revision) {
 		case 0:
sound/soc/codecs/wm_hubs.c +3 -0
index e93af7edd8f75..dd421e2fe7b21 100644
@@ -1223,6 +1223,9 @@ int wm_hubs_handle_analogue_pdata(struct snd_soc_component *component,
 		snd_soc_component_update_bits(component, WM8993_ADDITIONAL_CONTROL,
 				    WM8993_LINEOUT2_FB, WM8993_LINEOUT2_FB);
 
+	if (!hubs->micd_scthr)
+		return 0;
+
 	snd_soc_component_update_bits(component, WM8993_MICBIAS,
 			    WM8993_JD_SCTHR_MASK | WM8993_JD_THR_MASK |
 			    WM8993_MICB1_LVL | WM8993_MICB2_LVL,
sound/soc/codecs/wm_hubs.h +1 -0
index 4b8e5f0d6e32d..988b29e630607 100644
@@ -27,6 +27,7 @@ struct wm_hubs_data {
 	int hp_startup_mode;
 	int series_startup;
 	int no_series_update;
+	bool micd_scthr;
 
 	bool no_cache_dac_hp_direct;
 	struct list_head dcs_cache;
Signature · 1 line
2.25.1
author maintainer
#3 lore raw

PATCH 4/15 ASoC: wm8994: Ensure the device is resumed in wm89xx_mic_detect functions

In reply to Sasha Levin

From: Sylwester Nawrocki <[email protected]>

[ Upstream commit f5a2cda4f1db89776b64c4f0f2c2ac609527ac70 ]

When the wm8958_mic_detect, wm8994_mic_detect functions get called from the machine driver, e.g. from the card's late_probe() callback, the CODEC device may be PM runtime suspended and any regmap writes have no effect. Add PM runtime calls to these functions to ensure the device registers are updated as expected.
This suppresses an error during boot
"wm8994-codec: ASoC: error at snd_soc_component_update_bits on wm8994-codec" caused by the regmap access error due to the cache_only flag being set.

---
 sound/soc/codecs/wm8994.c | 8 ++++++++
 1 file changed, 8 insertions(+)
sound/soc/codecs/wm8994.c +8 -0
index 64635f9cdae65..6dbab3fc6537e 100644
@@ -3372,6 +3372,8 @@ int wm8994_mic_detect(struct snd_soc_component *component, struct snd_soc_jack *
 		return -EINVAL;
 	}
 
+	pm_runtime_get_sync(component->dev);
+
 	switch (micbias) {
 	case 1:
 		micdet = &wm8994->micdet[0];
@@ -3419,6 +3421,8 @@ int wm8994_mic_detect(struct snd_soc_component *component, struct snd_soc_jack *
 
 	snd_soc_dapm_sync(dapm);
 
+	pm_runtime_put(component->dev);
+
 	return 0;
 }
 EXPORT_SYMBOL_GPL(wm8994_mic_detect);
@@ -3786,6 +3790,8 @@ int wm8958_mic_detect(struct snd_soc_component *component, struct snd_soc_jack *
 		return -EINVAL;
 	}
 
+	pm_runtime_get_sync(component->dev);
+
 	if (jack) {
 		snd_soc_dapm_force_enable_pin(dapm, "CLK_SYS");
 		snd_soc_dapm_sync(dapm);
@@ -3854,6 +3860,8 @@ int wm8958_mic_detect(struct snd_soc_component *component, struct snd_soc_jack *
 		snd_soc_dapm_sync(dapm);
 	}
 
+	pm_runtime_put(component->dev);
+
 	return 0;
 }
 EXPORT_SYMBOL_GPL(wm8958_mic_detect);
Signature · 1 line
2.25.1
author maintainer
#4 lore raw

PATCH 5/15 ASoC: Intel: bytcr_rt5640: Add quirk for MPMAN Converter9 2-in-1

In reply to Sasha Levin

From: Hans de Goede <[email protected]>

[ Upstream commit 6a0137101f47301fff2da6ba4b9048383d569909 ]

The MPMAN Converter9 2-in-1 almost fully works with out default settings. The only problem is that it has only 1 speaker so any sounds only playing on the right channel get lost.

Add a quirk for this model using the default settings + MONO_SPEAKER.

---
 sound/soc/intel/boards/bytcr_rt5640.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
sound/soc/intel/boards/bytcr_rt5640.c +10 -0
index f7964d1ec486f..6012367f6fe48 100644
@@ -591,6 +591,16 @@ static const struct dmi_system_id byt_rt5640_quirk_table[] = {
 					BYT_RT5640_SSP0_AIF1 |
 					BYT_RT5640_MCLK_EN),
 	},
+	{	/* MPMAN Converter 9, similar hw as the I.T.Works TW891 2-in-1 */
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "MPMAN"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Converter9"),
+		},
+		.driver_data = (void *)(BYTCR_INPUT_DEFAULTS |
+					BYT_RT5640_MONO_SPEAKER |
+					BYT_RT5640_SSP0_AIF1 |
+					BYT_RT5640_MCLK_EN),
+	},
 	{
 		/* MPMAN MPWIN895CL */
 		.matches = {
Signature · 1 line
2.25.1
author maintainer
#5 lore raw

PATCH 6/15 RISC-V: Take text_mutex in ftrace_init_nop()

In reply to Sasha Levin

From: Palmer Dabbelt <[email protected]>

[ Upstream commit 66d18dbda8469a944dfec6c49d26d5946efba218 ]

Without this we get lockdep failures. They're spurious failures as SMP isn't up when ftrace_init_nop() is called. As far as I can tell the easiest fix is to just take the lock, which also seems like the safest fix.

---
 arch/riscv/include/asm/ftrace.h |  7 +++++++
 arch/riscv/kernel/ftrace.c      | 19 +++++++++++++++++++
 2 files changed, 26 insertions(+)
arch/riscv/include/asm/ftrace.h +7 -0
index c6dcc5291f972..02fbc175142e2 100644
@@ -63,4 +63,11 @@ do {									\
  * Let auipc+jalr be the basic *mcount unit*, so we make it 8 bytes here.
  */
 #define MCOUNT_INSN_SIZE 8
+
+#ifndef __ASSEMBLY__
+struct dyn_ftrace;
+int ftrace_init_nop(struct module *mod, struct dyn_ftrace *rec);
+#define ftrace_init_nop ftrace_init_nop
+#endif
+
 #endif
arch/riscv/kernel/ftrace.c +19 -0
index c40fdcdeb950a..291c579e12457 100644
@@ -88,6 +88,25 @@ int ftrace_make_nop(struct module *mod, struct dyn_ftrace *rec,
 	return __ftrace_modify_call(rec->ip, addr, false);
 }
 
+
+/*
+ * This is called early on, and isn't wrapped by
+ * ftrace_arch_code_modify_{prepare,post_process}() and therefor doesn't hold
+ * text_mutex, which triggers a lockdep failure.  SMP isn't running so we could
+ * just directly poke the text, but it's simpler to just take the lock
+ * ourselves.
+ */
+int ftrace_init_nop(struct module *mod, struct dyn_ftrace *rec)
+{
+	int out;
+
+	ftrace_arch_code_modify_prepare();
+	out = ftrace_make_nop(mod, rec, MCOUNT_ADDR);
+	ftrace_arch_code_modify_post_process();
+
+	return out;
+}
+
 int ftrace_update_ftrace_func(ftrace_func_t func)
 {
 	int ret = __ftrace_modify_call((unsigned long)&ftrace_call,
Signature · 1 line
2.25.1
author maintainer
#6 lore raw

PATCH 7/15 i2c: aspeed: Mask IRQ status to relevant bits

In reply to Sasha Levin

From: Eddie James <[email protected]>

[ Upstream commit 1a1d6db23ddacde0b15ea589e9103373e05af8de ]

Mask the IRQ status to only the bits that the driver checks. This prevents excessive driver warnings when operating in slave mode when additional bits are set that the driver doesn't handle.

---
 drivers/i2c/busses/i2c-aspeed.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/i2c/busses/i2c-aspeed.c +2 -0
index dad6e432de89f..bdcc3c9d0abe5 100644
@@ -69,6 +69,7 @@
  * These share bit definitions, so use the same values for the enable &
  * status bits.
  */
+#define ASPEED_I2CD_INTR_RECV_MASK			0xf000ffff
 #define ASPEED_I2CD_INTR_SDA_DL_TIMEOUT			BIT(14)
 #define ASPEED_I2CD_INTR_BUS_RECOVER_DONE		BIT(13)
 #define ASPEED_I2CD_INTR_SLAVE_MATCH			BIT(7)
@@ -604,6 +605,7 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id)
 	writel(irq_received & ~ASPEED_I2CD_INTR_RX_DONE,
 	       bus->base + ASPEED_I2C_INTR_STS_REG);
 	readl(bus->base + ASPEED_I2C_INTR_STS_REG);
+	irq_received &= ASPEED_I2CD_INTR_RECV_MASK;
 	irq_remaining = irq_received;
 
 #if IS_ENABLED(CONFIG_I2C_SLAVE)
Signature · 1 line
2.25.1
author maintainer
#7 lore raw

PATCH 8/15 s390/init: add missing __init annotations

In reply to Sasha Levin

From: Ilya Leoshkevich <[email protected]>

[ Upstream commit fcb2b70cdb194157678fb1a75f9ff499aeba3d2a ]

Add __init to reserve_memory_end, reserve_oldmem and remove_oldmem. Sometimes these functions are not inlined, and then the build complains about section mismatch.

---
 arch/s390/kernel/setup.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/s390/kernel/setup.c +3 -3
index 07b2b61a0289f..cc7726e962639 100644
@@ -634,7 +634,7 @@ static struct notifier_block kdump_mem_nb = {
 /*
  * Make sure that the area behind memory_end is protected
  */
-static void reserve_memory_end(void)
+static void __init reserve_memory_end(void)
 {
 	if (memory_end_set)
 		memblock_reserve(memory_end, ULONG_MAX);
@@ -643,7 +643,7 @@ static void reserve_memory_end(void)
 /*
  * Make sure that oldmem, where the dump is stored, is protected
  */
-static void reserve_oldmem(void)
+static void __init reserve_oldmem(void)
 {
 #ifdef CONFIG_CRASH_DUMP
 	if (OLDMEM_BASE)
@@ -655,7 +655,7 @@ static void reserve_oldmem(void)
 /*
  * Make sure that oldmem, where the dump is stored, is protected
  */
-static void remove_oldmem(void)
+static void __init remove_oldmem(void)
 {
 #ifdef CONFIG_CRASH_DUMP
 	if (OLDMEM_BASE)
Signature · 1 line
2.25.1
author maintainer
#8 lore raw

PATCH 9/15 lockdep: fix order in trace_hardirqs_off_caller()

In reply to Sasha Levin

From: Sven Schnelle <[email protected]>

[ Upstream commit 73ac74c7d489756d2313219a108809921dbfaea1 ]

Switch order so that locking state is consistent even if the IRQ tracer calls into lockdep again.

---
 kernel/trace/trace_preemptirq.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/trace/trace_preemptirq.c +2 -2
index 4d8e99fdbbbee..26b06b09c9f68 100644
@@ -63,14 +63,14 @@ NOKPROBE_SYMBOL(trace_hardirqs_on_caller);
 
 __visible void trace_hardirqs_off_caller(unsigned long caller_addr)
 {
+	lockdep_hardirqs_off(CALLER_ADDR0);
+
 	if (!this_cpu_read(tracing_irq_cpu)) {
 		this_cpu_write(tracing_irq_cpu, 1);
 		tracer_hardirqs_off(CALLER_ADDR0, caller_addr);
 		if (!in_nmi())
 			trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
 	}
-
-	lockdep_hardirqs_off(CALLER_ADDR0);
 }
 EXPORT_SYMBOL(trace_hardirqs_off_caller);
 NOKPROBE_SYMBOL(trace_hardirqs_off_caller);
Signature · 1 line
2.25.1
author maintainer
#9 lore raw

PATCH 10/15 EDAC/ghes: Check whether the driver is on the safe list correctly

In reply to Sasha Levin

From: Borislav Petkov <[email protected]>

[ Upstream commit 251c54ea26fa6029b01a76161a37a12fde5124e4 ]

With CONFIG_DEBUG_TEST_DRIVER_REMOVE=y, a system would try to probe, unregister and probe again a driver.

When ghes_edac is attempted to be loaded on a system which is not on the safe platforms list, ghes_edac_register() would return early. The unregister counterpart ghes_edac_unregister() would still attempt to unregister and exit early at the refcount test, leading to the refcount underflow below.

In order to not do *anything* on the unregister path too, reuse the force_load parameter and check it on that path too, before fumbling with the refcount.

ghes_edac: ghes_edac_register: entry
ghes_edac: ghes_edac_register: return -ENODEV

  ------------[ cut here ]------------
  refcount_t: underflow; use-after-free.
  WARNING: CPU: 10 PID: 1 at lib/refcount.c:28 refcount_warn_saturate+0xb9/0x100
  Modules linked in:
  CPU: 10 PID: 1 Comm: swapper/0 Not tainted 5.9.0-rc4+ #12
  Hardware name: GIGABYTE MZ01-CE1-00/MZ01-CE1-00, BIOS F02 08/29/2018
  RIP: 0010:refcount_warn_saturate+0xb9/0x100
  Code: 82 e8 fb 8f 4d 00 90 0f 0b 90 90 c3 80 3d 55 4c f5 00 00 75 88 c6 05 4c 4c f5 00 01 90 48 c7 c7 d0 8a 10 82 e8 d8 8f 4d 00 90 <0f> 0b 90 90 c3 80 3d 30 4c f5 00 00 0f 85 61 ff ff ff c6 05 23 4c
  RSP: 0018:ffffc90000037d58 EFLAGS: 00010292
  RAX: 0000000000000026 RBX: ffff88840b8da000 RCX: 0000000000000000
  RDX: 0000000000000001 RSI: ffffffff8216b24f RDI: 00000000ffffffff
  RBP: ffff88840c662e00 R08: 0000000000000001 R09: 0000000000000001
  R10: 0000000000000001 R11: 0000000000000046 R12: 0000000000000000
  R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000000
  FS:  0000000000000000(0000) GS:ffff88840ee80000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000000000000 CR3: 0000800002211000 CR4: 00000000003506e0
  Call Trace:
   ghes_edac_unregister
   ghes_remove
   platform_drv_remove
   really_probe
   driver_probe_device
   device_driver_attach
   __driver_attach
   ? device_driver_attach
   ? device_driver_attach
   bus_for_each_dev
   bus_add_driver
   driver_register
   ? bert_init
   ghes_init
   do_one_initcall
   ? rcu_read_lock_sched_held
   kernel_init_freeable
   ? rest_init
   kernel_init
   ret_from_fork
   ...
  ghes_edac: ghes_edac_unregister: FALSE, refcount: -1073741824
---
 drivers/edac/ghes_edac.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/edac/ghes_edac.c +4 -0
index 523dd56a798c9..0031819402d0c 100644
@@ -488,6 +488,7 @@ int ghes_edac_register(struct ghes *ghes, struct device *dev)
 		if (!force_load && idx < 0)
 			return -ENODEV;
 	} else {
+		force_load = true;
 		idx = 0;
 	}
 
@@ -586,6 +587,9 @@ void ghes_edac_unregister(struct ghes *ghes)
 	struct mem_ctl_info *mci;
 	unsigned long flags;
 
+	if (!force_load)
+		return;
+
 	mutex_lock(&ghes_reg_mutex);
 
 	if (!refcount_dec_and_test(&ghes_refcount))
Signature · 1 line
2.25.1
author maintainer
#10 lore raw

PATCH 11/15 drm/amdkfd: fix a memory leak issue

In reply to Sasha Levin

From: Dennis Li <[email protected]>

[ Upstream commit 087d764159996ae378b08c0fdd557537adfd6899 ]

In the resume stage of GPU recovery, start_cpsch will call pm_init which set pm->allocated as false, cause the next pm_release_ib has no chance to release ib memory.

Add pm_release_ib in stop_cpsch which will be called in the suspend stage of GPU recovery.

---
 drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +2 -0
index a2ed9c257cb0d..e9a2784400792 100644
@@ -1075,6 +1075,8 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 	unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	dqm_unlock(dqm);
 
+	pm_release_ib(&dqm->packets);
+
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
 	pm_uninit(&dqm->packets);
 
Signature · 6 lines
2.25.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
author maintainer
#11 lore raw

PATCH 12/15 drm/amd/display: update nv1x stutter latencies

In reply to Sasha Levin

From: Jun Lei <[email protected]>

[ Upstream commit c4790a8894232f39c25c7c546c06efe074e63384 ]

[why]
Recent characterization shows increased stutter latencies on some SKUs, leading to underflow.

[how]
Update SOC params to account for this worst case latency.

---
 drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c +2 -2
index 05b98eadc2899..46afd048c1ea3 100644
@@ -340,8 +340,8 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_0_nv14_soc = {
 			},
 		},
 	.num_states = 5,
-	.sr_exit_time_us = 8.6,
-	.sr_enter_plus_exit_time_us = 10.9,
+	.sr_exit_time_us = 11.6,
+	.sr_enter_plus_exit_time_us = 13.9,
 	.urgent_latency_us = 4.0,
 	.urgent_latency_pixel_data_only_us = 4.0,
 	.urgent_latency_pixel_mixed_with_vm_data_us = 4.0,
Signature · 6 lines
2.25.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
author maintainer
#12 lore raw

PATCH 13/15 drm/amdgpu/dc: Require primary plane to be enabled whenever the CRTC is

In reply to Sasha Levin

From: Michel Dänzer <[email protected]>

[ Upstream commit 2f228aab21bbc74e90e267a721215ec8be51daf7 ]

Don't check drm_crtc_state::active for this either, per its documentation in include/drm/drm_crtc.h:

* Hence drivers must not consult @active in their various
* &drm_mode_config_funcs.atomic_check callback to reject an atomic
* commit.

atomic_remove_fb disables the CRTC as needed for disabling the primary plane.

This prevents at least the following problems if the primary plane gets disabled (e.g. due to destroying the FB assigned to the primary plane, as happens e.g. with mutter in Wayland mode):

* The legacy cursor ioctl returned EINVAL for a non-0 cursor FB ID (which enables the cursor plane).
* If the cursor plane was enabled, changing the legacy DPMS property value from off to on returned EINVAL.

v2:
* Minor changes to code comment and commit log, per review feedback.

GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1108
GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1165
GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1344

---
 .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 32 ++++++-------------
 1 file changed, 10 insertions(+), 22 deletions(-)
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +10 -22
index 60e50181f6d39..2384aa018993d 100644
@@ -4299,19 +4299,6 @@ static void dm_crtc_helper_disable(struct drm_crtc *crtc)
 {
 }
 
-static bool does_crtc_have_active_cursor(struct drm_crtc_state *new_crtc_state)
-{
-	struct drm_device *dev = new_crtc_state->crtc->dev;
-	struct drm_plane *plane;
-
-	drm_for_each_plane_mask(plane, dev, new_crtc_state->plane_mask) {
-		if (plane->type == DRM_PLANE_TYPE_CURSOR)
-			return true;
-	}
-
-	return false;
-}
-
 static int count_crtc_active_planes(struct drm_crtc_state *new_crtc_state)
 {
 	struct drm_atomic_state *state = new_crtc_state->state;
@@ -4391,19 +4378,20 @@ static int dm_crtc_helper_atomic_check(struct drm_crtc *crtc,
 		return ret;
 	}
 
-	/* In some use cases, like reset, no stream is attached */
-	if (!dm_crtc_state->stream)
-		return 0;
-
 	/*
-	 * We want at least one hardware plane enabled to use
-	 * the stream with a cursor enabled.
+	 * We require the primary plane to be enabled whenever the CRTC is, otherwise
+	 * drm_mode_cursor_universal may end up trying to enable the cursor plane while all other
+	 * planes are disabled, which is not supported by the hardware. And there is legacy
+	 * userspace which stops using the HW cursor altogether in response to the resulting EINVAL.
 	 */
-	if (state->enable && state->active &&
-	    does_crtc_have_active_cursor(state) &&
-	    dm_crtc_state->active_planes == 0)
+	if (state->enable &&
+	    !(state->plane_mask & drm_plane_mask(crtc->primary)))
 		return -EINVAL;
 
+	/* In some use cases, like reset, no stream is attached */
+	if (!dm_crtc_state->stream)
+		return 0;
+
 	if (dc_validate_stream(dc, dm_crtc_state->stream) == DC_OK)
 		return 0;
 
Signature · 6 lines
2.25.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#13 lore raw

Re: [PATCH AUTOSEL 5.4 13/15] drm/amdgpu/dc: Require primary plane to be enabled whenever the CRTC is

Quoted 29 lines · Sasha Levin

From: Michel Dänzer <[email protected]>

[ Upstream commit 2f228aab21bbc74e90e267a721215ec8be51daf7 ]

Don't check drm_crtc_state::active for this either, per its documentation in include/drm/drm_crtc.h:

* Hence drivers must not consult @active in their various
* &drm_mode_config_funcs.atomic_check callback to reject an atomic
* commit.

atomic_remove_fb disables the CRTC as needed for disabling the primary plane.

This prevents at least the following problems if the primary plane gets disabled (e.g. due to destroying the FB assigned to the primary plane, as happens e.g. with mutter in Wayland mode):

* The legacy cursor ioctl returned EINVAL for a non-0 cursor FB ID (which enables the cursor plane).
* If the cursor plane was enabled, changing the legacy DPMS property value from off to on returned EINVAL.

v2:
* Minor changes to code comment and commit log, per review feedback.

GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1108
GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1165
GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1344

Quoted 6 lines

I'm a bit nervous about this getting backported so far back so quickly. I'd prefer waiting for 5.9 final first at least.

Signature · 6 lines
Earthling Michel Dänzer               |               https://redhat.com
Libre software enthusiast             |             Mesa and X developer
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
author maintainer
#14 lore raw

Re: [PATCH AUTOSEL 5.4 13/15] drm/amdgpu/dc: Require primary plane to be enabled whenever the CRTC is

Quoted 37 lines · Michel Dänzer
Quoted 29 lines · Sasha Levin

From: Michel Dänzer <[email protected]>

[ Upstream commit 2f228aab21bbc74e90e267a721215ec8be51daf7 ]

Don't check drm_crtc_state::active for this either, per its documentation in include/drm/drm_crtc.h:

* Hence drivers must not consult @active in their various
* &drm_mode_config_funcs.atomic_check callback to reject an atomic
* commit.

atomic_remove_fb disables the CRTC as needed for disabling the primary plane.

This prevents at least the following problems if the primary plane gets disabled (e.g. due to destroying the FB assigned to the primary plane, as happens e.g. with mutter in Wayland mode):

* The legacy cursor ioctl returned EINVAL for a non-0 cursor FB ID (which enables the cursor plane).
* If the cursor plane was enabled, changing the legacy DPMS property value from off to on returned EINVAL.

v2:
* Minor changes to code comment and commit log, per review feedback.

GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1108
GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1165
GitLab: https://gitlab.gnome.org/GNOME/mutter/-/issues/1344

Quoted 6 lines
Quoted 1 line

I'm a bit nervous about this getting backported so far back so quickly. I'd prefer waiting for 5.9 final first at least.

Will drop it for now, thanks.

Signature · 2 lines
Thanks,
Sasha
author maintainer
#15 lore raw

PATCH 14/15 i2c: core: Call i2c_acpi_install_space_handler() before i2c_acpi_register_devices()

In reply to Sasha Levin

From: Hans de Goede <[email protected]>

[ Upstream commit 21653a4181ff292480599dad996a2b759ccf050f ]

Some ACPI i2c-devices _STA method (which is used to detect if the device is present) use autodetection code which probes which device is present over i2c. This requires the I2C ACPI OpRegion handler to be registered before we enumerate i2c-clients under the i2c-adapter.

This fixes the i2c touchpad on the Lenovo ThinkBook 14-IIL and ThinkBook 15 IIL not getting an i2c-client instantiated and thus not working.

---
 drivers/i2c/i2c-core-base.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/i2c-core-base.c +1 -1
index def62d5b42ca7..2dfe2ffcf8825 100644
@@ -1385,8 +1385,8 @@ static int i2c_register_adapter(struct i2c_adapter *adap)
 
 	/* create pre-declared device nodes */
 	of_i2c_register_devices(adap);
-	i2c_acpi_register_devices(adap);
 	i2c_acpi_install_space_handler(adap);
+	i2c_acpi_register_devices(adap);
 
 	if (adap->nr < __i2c_first_dynamic_bus_num)
 		i2c_scan_static_board_info(adap);
Signature · 1 line
2.25.1
author maintainer
#16 lore raw

PATCH 15/15 objtool: Fix noreturn detection for ignored functions

In reply to Sasha Levin

From: Josh Poimboeuf <[email protected]>

[ Upstream commit db6c6a0df840e3f52c84cc302cc1a08ba11a4416 ]

When a function is annotated with STACK_FRAME_NON_STANDARD, objtool doesn't validate its code paths. It also skips sibling call detection within the function.

But sibling call detection is actually needed for the case where the ignored function doesn't have any return instructions. Otherwise objtool naively marks the function as implicit static noreturn, which affects the reachability of its callers, resulting in "unreachable instruction" warnings.

Fix it by just enabling sibling call detection for ignored functions. The 'insn->ignore' check in add_jump_destinations() is no longer needed after

e6da9567959e ("objtool: Don't use ignore flag for fake jumps").

Fixes the following warning:

arch/x86/kvm/vmx/vmx.o: warning: objtool: vmx_handle_exit_irqoff()+0x142: unreachable instruction

which triggers on an allmodconfig with CONFIG_GCOV_KERNEL unset.

---
 tools/objtool/check.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/objtool/check.c +1 -1
index 48b234d8f251e..1b7e748170e54 100644
@@ -556,7 +556,7 @@ static int add_jump_destinations(struct objtool_file *file)
 		    insn->type != INSN_JUMP_UNCONDITIONAL)
 			continue;
 
-		if (insn->ignore || insn->offset == FAKE_JUMP_OFFSET)
+		if (insn->offset == FAKE_JUMP_OFFSET)
 			continue;
 
 		rela = find_rela_by_dest_range(insn->sec, insn->offset,
Signature · 1 line
2.25.1