Linux 内核模块 - 打开创建的字符设备导致错误
Linux kernel module - opening created char device results in a bug
我刚开始阅读 Ldd3 并尝试实现一个简单的模块，或多或少遵循书中的示例。模块将具有自定义 fops（空函数）的字符设备添加到内核。模块加载没有问题，但是当我尝试打开设备时，进程被终止并且 dmesg 打印出调试信息。我的内核是 4.8.0。 x86-64。这是我的代码
#include <linux/init.h>
#include <linux/module.h>
#include <linux/types.h> // for dev_t
#include <linux/kdev_t.h> // for MAJOR, MINOR macros etc..
#include <linux/fs.h> // for alloc_chrdev_region etc ..
#include <linux/cdev.h> // to register char devic
MODULE_LICENSE("Dual BSD/GPL");
dev_t hello_num;
struct cdev *my_cdev;
loff_t hello_llseek(struct file *filepo, loff_t offset, int a){
return 0;
}
int hello_open(struct inode *inode, struct file *flip){
printk(KERN_ALERT "you opened this device");
return 0;
}
ssize_t hello_read(struct file *filep, char __user *ptr, size_t sizee, loff_t *gela){
return 0;
}
ssize_t hello_write(struct file *filep, const char __user *ptr, size_t sizee, loff_t *gela){
return 0;
}
int hello_release(struct inode *inode, struct file *filep){
return 0;
}
static int hello_init(void)
{
int cdev_add_return = 999; //DEBUG
struct file_operations hello_fops = {
.owner = THIS_MODULE,
.llseek = hello_llseek,
.read = hello_read,
.write = hello_write,
.open = hello_open,
.release = hello_release,
};
alloc_chrdev_region(&hello_num, 0, 3, "hello");
my_cdev = cdev_alloc();
if(my_cdev == 0){
printk(KERN_DEBUG "my_cdev is NULL");
}
printk(KERN_DEBUG "mc_cdev address: %p", my_cdev); // DEBUG
my_cdev->ops = &hello_fops;
if((cdev_add_return = cdev_add(my_cdev, hello_num, 3)) != 0){
printk(KERN_DEBUG "cdev_add errrored \n");
}
printk(KERN_DEBUG "cdev_add returned: %d", cdev_add_return); // DEBUG
printk(KERN_INFO "Hello World: major num: %ld \n", (long) MAJOR(hello_num));
return 0;
}
static void hello_exit(void)
{
cdev_del(my_cdev);
unregister_chrdev_region(hello_num, 3);
printk(KERN_INFO "Goodbye\n");
}
module_init(hello_init);
module_exit(hello_exit);
调试信息如下：
[ 3504.756954] mc_cdev address: ffff8800340ed280
[ 3504.756967] cdev_add returned: 0
[ 3504.756975] Hello World: major num: 245
[ 3512.356621] BUG: unable to handle kernel paging request at ffffffff810b3d1d
[ 3512.359676] IP: [<ffffffff811029f1>] try_module_get+0x41/0x110
[ 3512.359676] PGD 1809067 PUD 180a063 PMD 10001e1
[ 3512.359676] Oops: 0003 [#2] PREEMPT SMP
[ 3512.359676] Modules linked in: hello(O) joydev mousedev arc4 ath9k ath9k_common ath9k_hw ath snd_hda_codec_realtek input_leds psmouse mac_hid uvcvideo mac80211 snd_hda_codec_generic evdev serio_raw iTCO_wdt atkbd iTCO_vendor_support libps2 videobuf2_vmalloc videobuf2_memops videobuf2_v4l2 videobuf2_core asus_wmi snd_hda_intel videodev snd_hda_codec media i915 coretemp lpc_ich i2c_i801 pcspkr i2c_smbus i8042 drm_kms_helper atl1c cfg80211 snd_hda_core drm sparse_keymap wmi led_class rfkill serio snd_hwdep snd_pcm syscopyarea thermal sysfillrect snd_timer fjes sysimgblt ac fb_sys_fops shpchp snd intel_agp intel_gtt battery i2c_algo_bit soundcore video button acpi_cpufreq tpm_tis tpm_tis_core tpm squashfs loop sch_fq_codel ip_tables x_tables ext4 crc16 jbd2 mbcache uas usb_storage sd_mod uhci_hcd ahci
[ 3512.359676] libahci libata scsi_mod ehci_pci ehci_hcd usbcore usb_common [last unloaded: hello]
[ 3512.359676] CPU: 1 PID: 2438 Comm: checkout Tainted: G R D O 4.8.0-ARCH #1
[ 3512.359676] Hardware name: ASUSTeK Computer INC. 1005P/1005P, BIOS 0706 01/04/2010
[ 3512.359676] task: ffff880034035b00 task.stack: ffff8800341a8000
[ 3512.359676] RIP: 0010:[<ffffffff811029f1>] [<ffffffff811029f1>] try_module_get+0x41/0x110
[ 3512.359676] RSP: 0018:ffff8800341abc30 EFLAGS: 00010206
[ 3512.359676] RAX: 0000000048000009 RBX: ffff8800340ed280 RCX: 0000000048000009
[ 3512.359676] RDX: 000000004800000a RSI: ffffffff810b3d1d RDI: ffffffff810b39e5
[ 3512.359676] RBP: ffff8800341abc48 R08: 0000000000000000 R09: 0000000000000000
[ 3512.359676] R10: 0000000000000000 R11: ffff88002d65df38 R12: ffff880034064320
[ 3512.469705] R13: ffff88003834f600 R14: ffff8800340ed280 R15: 0000000000000000
[ 3512.469705] FS: 00007f5bd10a9400(0000) GS:ffff88003f500000(0000) knlGS:0000000000000000
[ 3512.469705] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 3512.469705] CR2: ffffffff810b3d1d CR3: 000000003d315000 CR4: 00000000000006e0
[ 3512.469705] Stack:
[ 3512.469705] ffff8800340ed280 ffff880034064320 ffff88003834f600 ffff8800341abc98
[ 3512.469705] ffffffff8120c99d ffff880034064320 0000000000008000 0000000064a8303a
[ 3512.469705] ffff88003834f600 ffff8800341afcc0 ffff880034064320 0000000000000000
[ 3512.469705] Call Trace:
[ 3512.469705] [<ffffffff8120c99d>] chrdev_open+0x7d/0x1e0
[ 3512.469705] [<ffffffff812052c5>] do_dentry_open+0x205/0x2e0
[ 3512.469705] [<ffffffff8120c920>] ? cdev_put+0x30/0x30
[ 3512.469705] [<ffffffff8120637c>] vfs_open+0x4c/0x70
[ 3512.469705] [<ffffffff81213fbb>] ? may_open+0x9b/0x100
[ 3512.469705] [<ffffffff81216332>] path_openat+0x282/0x1170
[ 3512.469705] [<ffffffff811f32b9>] ? unlock_page_memcg+0x29/0x60
[ 3512.469705] [<ffffffff811bdbcb>] ? page_add_file_rmap+0x5b/0x140
[ 3512.469705] [<ffffffff81176de3>] ? filemap_map_pages+0x233/0x410
[ 3512.469705] [<ffffffff812183f1>] do_filp_open+0x91/0x100
[ 3512.469705] [<ffffffff81226129>] ? __alloc_fd+0xc9/0x180
[ 3512.469705] [<ffffffff812067a7>] do_sys_open+0x147/0x210
[ 3512.469705] [<ffffffff8120688e>] SyS_open+0x1e/0x20
[ 3512.469705] [<ffffffff815f7072>] entry_SYSCALL_64_fastpath+0x1a/0xa4
[ 3512.469705] Code: 53 65 ff 05 c2 a8 f0 7e 83 3f 02 0f 84 bd 00 00 00 8b 8f 38 03 00 00 85 c9 0f 84 af 00 00 00 8d 51 01 48 8d b7 38 03 00 00 89 c8 <f0> 0f b1 97 38 03 00 00 39 c8 89 c2 75 7e 4c 8b 6d 08 0f 1f 44
[ 3512.469705] RIP [<ffffffff811029f1>] try_module_get+0x41/0x110
[ 3512.469705] RSP <ffff8800341abc30>
[ 3512.469705] CR2: ffffffff810b3d1d
[ 3512.469705] ---[ end trace c8094481edecca95 ]---
生成文件：
obj-m += hello.o
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
clean:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean
我手动创建设备节点。谁能指出问题出在哪里？
提前致谢
您的具有文件操作的结构在函数的本地声明。因此，当 hello_init 完成时，指向该结构的指针不再有效。
当您打开字符设备时，内核会尝试查看该结构，但它不可访问。
常见的做法是用文件操作声明结构静态：
static struct file_operations hello_fops = {...};
我刚开始阅读 Ldd3 并尝试实现一个简单的模块，或多或少遵循书中的示例。模块将具有自定义 fops（空函数）的字符设备添加到内核。模块加载没有问题，但是当我尝试打开设备时，进程被终止并且 dmesg 打印出调试信息。我的内核是 4.8.0。 x86-64。这是我的代码
#include <linux/init.h>
#include <linux/module.h>
#include <linux/types.h> // for dev_t
#include <linux/kdev_t.h> // for MAJOR, MINOR macros etc..
#include <linux/fs.h> // for alloc_chrdev_region etc ..
#include <linux/cdev.h> // to register char devic
MODULE_LICENSE("Dual BSD/GPL");
dev_t hello_num;
struct cdev *my_cdev;
loff_t hello_llseek(struct file *filepo, loff_t offset, int a){
return 0;
}
int hello_open(struct inode *inode, struct file *flip){
printk(KERN_ALERT "you opened this device");
return 0;
}
ssize_t hello_read(struct file *filep, char __user *ptr, size_t sizee, loff_t *gela){
return 0;
}
ssize_t hello_write(struct file *filep, const char __user *ptr, size_t sizee, loff_t *gela){
return 0;
}
int hello_release(struct inode *inode, struct file *filep){
return 0;
}
static int hello_init(void)
{
int cdev_add_return = 999; //DEBUG
struct file_operations hello_fops = {
.owner = THIS_MODULE,
.llseek = hello_llseek,
.read = hello_read,
.write = hello_write,
.open = hello_open,
.release = hello_release,
};
alloc_chrdev_region(&hello_num, 0, 3, "hello");
my_cdev = cdev_alloc();
if(my_cdev == 0){
printk(KERN_DEBUG "my_cdev is NULL");
}
printk(KERN_DEBUG "mc_cdev address: %p", my_cdev); // DEBUG
my_cdev->ops = &hello_fops;
if((cdev_add_return = cdev_add(my_cdev, hello_num, 3)) != 0){
printk(KERN_DEBUG "cdev_add errrored \n");
}
printk(KERN_DEBUG "cdev_add returned: %d", cdev_add_return); // DEBUG
printk(KERN_INFO "Hello World: major num: %ld \n", (long) MAJOR(hello_num));
return 0;
}
static void hello_exit(void)
{
cdev_del(my_cdev);
unregister_chrdev_region(hello_num, 3);
printk(KERN_INFO "Goodbye\n");
}
module_init(hello_init);
module_exit(hello_exit);
调试信息如下：
[ 3504.756954] mc_cdev address: ffff8800340ed280
[ 3504.756967] cdev_add returned: 0
[ 3504.756975] Hello World: major num: 245
[ 3512.356621] BUG: unable to handle kernel paging request at ffffffff810b3d1d
[ 3512.359676] IP: [<ffffffff811029f1>] try_module_get+0x41/0x110
[ 3512.359676] PGD 1809067 PUD 180a063 PMD 10001e1
[ 3512.359676] Oops: 0003 [#2] PREEMPT SMP
[ 3512.359676] Modules linked in: hello(O) joydev mousedev arc4 ath9k ath9k_common ath9k_hw ath snd_hda_codec_realtek input_leds psmouse mac_hid uvcvideo mac80211 snd_hda_codec_generic evdev serio_raw iTCO_wdt atkbd iTCO_vendor_support libps2 videobuf2_vmalloc videobuf2_memops videobuf2_v4l2 videobuf2_core asus_wmi snd_hda_intel videodev snd_hda_codec media i915 coretemp lpc_ich i2c_i801 pcspkr i2c_smbus i8042 drm_kms_helper atl1c cfg80211 snd_hda_core drm sparse_keymap wmi led_class rfkill serio snd_hwdep snd_pcm syscopyarea thermal sysfillrect snd_timer fjes sysimgblt ac fb_sys_fops shpchp snd intel_agp intel_gtt battery i2c_algo_bit soundcore video button acpi_cpufreq tpm_tis tpm_tis_core tpm squashfs loop sch_fq_codel ip_tables x_tables ext4 crc16 jbd2 mbcache uas usb_storage sd_mod uhci_hcd ahci
[ 3512.359676] libahci libata scsi_mod ehci_pci ehci_hcd usbcore usb_common [last unloaded: hello]
[ 3512.359676] CPU: 1 PID: 2438 Comm: checkout Tainted: G R D O 4.8.0-ARCH #1
[ 3512.359676] Hardware name: ASUSTeK Computer INC. 1005P/1005P, BIOS 0706 01/04/2010
[ 3512.359676] task: ffff880034035b00 task.stack: ffff8800341a8000
[ 3512.359676] RIP: 0010:[<ffffffff811029f1>] [<ffffffff811029f1>] try_module_get+0x41/0x110
[ 3512.359676] RSP: 0018:ffff8800341abc30 EFLAGS: 00010206
[ 3512.359676] RAX: 0000000048000009 RBX: ffff8800340ed280 RCX: 0000000048000009
[ 3512.359676] RDX: 000000004800000a RSI: ffffffff810b3d1d RDI: ffffffff810b39e5
[ 3512.359676] RBP: ffff8800341abc48 R08: 0000000000000000 R09: 0000000000000000
[ 3512.359676] R10: 0000000000000000 R11: ffff88002d65df38 R12: ffff880034064320
[ 3512.469705] R13: ffff88003834f600 R14: ffff8800340ed280 R15: 0000000000000000
[ 3512.469705] FS: 00007f5bd10a9400(0000) GS:ffff88003f500000(0000) knlGS:0000000000000000
[ 3512.469705] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 3512.469705] CR2: ffffffff810b3d1d CR3: 000000003d315000 CR4: 00000000000006e0
[ 3512.469705] Stack:
[ 3512.469705] ffff8800340ed280 ffff880034064320 ffff88003834f600 ffff8800341abc98
[ 3512.469705] ffffffff8120c99d ffff880034064320 0000000000008000 0000000064a8303a
[ 3512.469705] ffff88003834f600 ffff8800341afcc0 ffff880034064320 0000000000000000
[ 3512.469705] Call Trace:
[ 3512.469705] [<ffffffff8120c99d>] chrdev_open+0x7d/0x1e0
[ 3512.469705] [<ffffffff812052c5>] do_dentry_open+0x205/0x2e0
[ 3512.469705] [<ffffffff8120c920>] ? cdev_put+0x30/0x30
[ 3512.469705] [<ffffffff8120637c>] vfs_open+0x4c/0x70
[ 3512.469705] [<ffffffff81213fbb>] ? may_open+0x9b/0x100
[ 3512.469705] [<ffffffff81216332>] path_openat+0x282/0x1170
[ 3512.469705] [<ffffffff811f32b9>] ? unlock_page_memcg+0x29/0x60
[ 3512.469705] [<ffffffff811bdbcb>] ? page_add_file_rmap+0x5b/0x140
[ 3512.469705] [<ffffffff81176de3>] ? filemap_map_pages+0x233/0x410
[ 3512.469705] [<ffffffff812183f1>] do_filp_open+0x91/0x100
[ 3512.469705] [<ffffffff81226129>] ? __alloc_fd+0xc9/0x180
[ 3512.469705] [<ffffffff812067a7>] do_sys_open+0x147/0x210
[ 3512.469705] [<ffffffff8120688e>] SyS_open+0x1e/0x20
[ 3512.469705] [<ffffffff815f7072>] entry_SYSCALL_64_fastpath+0x1a/0xa4
[ 3512.469705] Code: 53 65 ff 05 c2 a8 f0 7e 83 3f 02 0f 84 bd 00 00 00 8b 8f 38 03 00 00 85 c9 0f 84 af 00 00 00 8d 51 01 48 8d b7 38 03 00 00 89 c8 <f0> 0f b1 97 38 03 00 00 39 c8 89 c2 75 7e 4c 8b 6d 08 0f 1f 44
[ 3512.469705] RIP [<ffffffff811029f1>] try_module_get+0x41/0x110
[ 3512.469705] RSP <ffff8800341abc30>
[ 3512.469705] CR2: ffffffff810b3d1d
[ 3512.469705] ---[ end trace c8094481edecca95 ]---
生成文件：
obj-m += hello.o
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
clean:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean
我手动创建设备节点。谁能指出问题出在哪里？ 提前致谢
您的具有文件操作的结构在函数的本地声明。因此，当 hello_init 完成时，指向该结构的指针不再有效。
当您打开字符设备时，内核会尝试查看该结构，但它不可访问。
常见的做法是用文件操作声明结构静态：
static struct file_operations hello_fops = {...};