Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Jia-Ju Bai <baijiaju1990@gmail.com>
Subject: [BUG] Possible sleep-in-atomic-context bugs involving
 regmap_lock_mutex()
To:     broonie@kernel.org, gregkh@linuxfoundation.org, rafael@kernel.org,
        songjun.wu@microchip.com, architt@codeaurora.org
Cc:     linux-kernel@vger.kernel.org
Message-ID: <ffbd81ae-fd30-dc57-c3ac-463f7fadb9ac@gmail.com>
Date:   Tue, 28 Aug 2018 16:57:51 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101
 Thunderbird/52.2.0
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Content-Language: en-US
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hello,

My static tool DSAC reports many sleep-in-atomic-context bugs involving 
regmap_lock_mutex(), so I wonder whether this function is possible to be 
executed in atomic context.

Here are some example bugs and their call paths in Linux-4.16 (from 
bottom to top, and [FUNC_PTR] means that there is a function pointer call):

[FUNC] mutex_lock_nested
drivers/base/regmap/regmap.c, 468:
     mutex_lock_nested in regmap_lock_mutex
drivers/base/regmap/regmap.c, 2503:
     [FUNC_PTR]regmap_lock_mutex in regmap_read
drivers/clk/clk-aspeed.c, 215:
     regmap_read in aspeed_clk_is_enabled
drivers/clk/clk-aspeed.c, 230:
     aspeed_clk_is_enabled in aspeed_clk_enable
drivers/clk/clk-aspeed.c, 228:
     _raw_spin_lock_irqsave in aspeed_clk_enable

[FUNC] mutex_lock_nested
drivers/base/regmap/regmap.c, 468:
     mutex_lock_nested in regmap_lock_mutex
drivers/base/regmap/regmap.c, 2821:
     [FUNC_PTR]regmap_lock_mutex in regmap_update_bits_base
drivers/clk/clk-aspeed.c, 270:
     regmap_update_bits_base in aspeed_clk_disable
drivers/clk/clk-aspeed.c, 267:
     _raw_spin_lock_irqsave in aspeed_clk_disable

[FUNC] mutex_lock_nested
drivers/base/regmap/regmap.c, 468:
     mutex_lock_nested in regmap_lock_mutex
drivers/base/regmap/regmap.c, 2503:
     [FUNC_PTR]regmap_lock_mutex in regmap_read
drivers/char/ipmi/bt-bmc.c, 385:
     regmap_read in bt_bmc_irq

[FUNC] mutex_lock_nested
drivers/base/regmap/regmap.c, 468:
     mutex_lock_nested in regmap_lock_mutex
drivers/base/regmap/regmap.c, 2821:
     [FUNC_PTR]regmap_lock_mutex in regmap_update_bits_base
drivers/gpu/drm/bridge/synopsys/dw-hdmi.c, 213:
     regmap_update_bits_base in hdmi_modb
drivers/gpu/drm/bridge/synopsys/dw-hdmi.c, 429:
     hdmi_modb in hdmi_set_cts_n
drivers/gpu/drm/bridge/synopsys/dw-hdmi.c, 527:
     hdmi_set_cts_n in hdmi_set_clk_regenerator
drivers/gpu/drm/bridge/synopsys/dw-hdmi.c, 524:
     spin_lock_irq in hdmi_set_clk_regenerator

[FUNC] mutex_lock_nested
drivers/base/regmap/regmap.c, 468:
     mutex_lock_nested in regmap_lock_mutex
drivers/base/regmap/regmap.c, 2503:
     [FUNC_PTR]regmap_lock_mutex in regmap_read
drivers/media/platform/atmel/atmel-isc.c, 1603:
     regmap_read in isc_interrupt (interrupt handler)

I find the code about the assignment of regmap_lock_mutex():
         if ((bus && bus->fast_io) ||
             config->fast_io) {
             spin_lock_init(&map->spinlock);
             map->lock = regmap_lock_spinlock;
             map->unlock = regmap_unlock_spinlock;
             lockdep_set_class_and_name(&map->spinlock,
                            lock_key, lock_name);
         } else {
             mutex_init(&map->mutex);
             map->lock = regmap_lock_mutex;
             map->unlock = regmap_unlock_mutex;
             lockdep_set_class_and_name(&map->mutex,
                            lock_key, lock_name);
         }

But after reading the code, I cannot find the relationship between the 
if condition and atomic context.

I am looking forward to your reply, thanks in advance :)


Best wishes,
Jia-Ju Bai