nfsv3挂载参数带soft或actimeo=1能挂载成功，不带这两个参数时挂载hung住:
mount -t nfs -o vers=3 ${server_ip}:/svr/export /mnt # hung
mount -t nfs -o vers=3,soft ${server_ip}:/svr/export /mnt # 成功
mount -t nfs -o vers=3,actimeo=1 ${server_ip}:/svr/export /mnt # 成功类似问题请参考Bug 204395 - NFS v3 mount hung。
以下脚本验证都会走到nfs_fs_mount_common():
kprobe_func_name=nfs_fs_mount_common
cd /sys/kernel/debug/tracing/
# 可以用 kprobe 跟踪的函数
cat available_filter_functions | grep ${kprobe_func_name}
echo 1 > tracing_on
echo "p:p_${kprobe_func_name} ${kprobe_func_name}" >> kprobe_events
echo 1 > events/kprobes/p_${kprobe_func_name}/enable
# echo stacktrace > events/kprobes/p_${kprobe_func_name}/trigger # 打印栈
# echo '!stacktrace' > events/kprobes/p_${kprobe_func_name}/trigger # 关闭栈
# echo 0 > events/kprobes/p_${kprobe_func_name}/enable
# echo "-:p_${kprobe_func_name}" >> kprobe_events
echo 0 > trace # 清除trace信息
cat trace_pipe脚本查看《nfs调试方法》。
vim删除某个重复的栈:
:%s/\[<0>\] grab_super+0x2b\/0x90\_.\{-}=======================================================//g
:%s/\[<0>\] iterate_supers+0x7f\/0x100\_.\{-}=======================================================//g找到以下可疑的nfs相关的栈:
=============== 进程 1232699 线程 1232699 kworker/u256:0+flush-0:48 栈信息 ===============
[<0>] inode_wait_for_writeback+0x21/0x30
[<0>] evict+0xbc/0x1a0
[<0>] __dentry_kill+0xdd/0x180
[<0>] dentry_kill+0x4d/0x260
[<0>] dput+0x183/0x200
[<0>] __put_nfs_open_context+0xd0/0x130 [nfs]
[<0>] nfs_free_request+0xb7/0x180 [nfs]
[<0>] nfs_release_request+0x59/0x80 [nfs]
[<0>] nfs_do_writepage+0x1bf/0x2d0 [nfs]
[<0>] nfs_writepages_callback+0xf/0x20 [nfs]
[<0>] write_cache_pages+0x187/0x410
[<0>] nfs_writepages+0xb0/0x170 [nfs]
[<0>] do_writepages+0x4b/0xe0
[<0>] __writeback_single_inode+0x3d/0x330
[<0>] writeback_sb_inodes+0x1ad/0x4b0
[<0>] __writeback_inodes_wb+0x5d/0xb0
[<0>] wb_writeback+0x26c/0x300
[<0>] wb_workfn+0x1dc/0x4c0
[<0>] process_one_work+0x195/0x3d0
[<0>] worker_thread+0x30/0x390
[<0>] kthread+0x113/0x130
[<0>] ret_from_fork+0x35/0x40
=======================================================rpm2cpio kernel-debuginfo-4.19.90-23.16.v2101.ky10.x86_64.rpm | cpio -div
./klinux-4.19/scripts/faddr2line usr/lib/debug/lib/modules/4.19.90-23.16.v2101.ky10.x86_64/kernel/fs/nfs/nfs.ko.debug nfs_release_request+0x59/0x80
nfs_release_request+0x59/0x80:
nfs_page_group_destroy at /usr/src/debug/kernel-4.19.90/linux-4.19.90-23.16.v2101.ky10.x86_64/fs/nfs/pagelist.c:314
(inlined by) kref_put at /usr/src/debug/kernel-4.19.90/linux-4.19.90-23.16.v2101.ky10.x86_64/./include/linux/kref.h:70
(inlined by) nfs_release_request at /usr/src/debug/kernel-4.19.90/linux-4.19.90-23.16.v2101.ky10.x86_64/fs/nfs/pagelist.c:458
./klinux-4.19/scripts/faddr2line usr/lib/debug/lib/modules/4.19.90-23.16.v2101.ky10.x86_64/kernel/fs/nfs/nfs.ko.debug nfs_do_writepage+0x1bf/0x2d0
nfs_do_writepage+0x1bf/0x2d0:
nfs_do_writepage at /usr/src/debug/kernel-4.19.90/linux-4.19.90-23.16.v2101.ky10.x86_64/fs/nfs/write.c:679nfs_do_writepage内联了太多函数，所以要查看反汇编:
# fs/nfs/write.c: 674 if (ret == -EAGAIN) {
0xffffffffc08b2600 <nfs_do_writepage+0x1a0>: cmp $0xfffffff5,%r14d
0xffffffffc08b2604 <nfs_do_writepage+0x1a4>: je 0xffffffffc08b2671 <nfs_do_writepage+0x211>
# fs/nfs/write.c: 679 return ret;
0xffffffffc08b2606 <nfs_do_writepage+0x1a6>: pop %rbx
0xffffffffc08b2607 <nfs_do_writepage+0x1a7>: mov %r14d,%eax
0xffffffffc08b260a <nfs_do_writepage+0x1aa>: pop %rbp
0xffffffffc08b260b <nfs_do_writepage+0x1ab>: pop %r12
0xffffffffc08b260d <nfs_do_writepage+0x1ad>: pop %r13
0xffffffffc08b260f <nfs_do_writepage+0x1af>: pop %r14
0xffffffffc08b2611 <nfs_do_writepage+0x1b1>: pop %r15
0xffffffffc08b2613 <nfs_do_writepage+0x1b3>: retq
# fs/nfs/write.c: 663 nfs_page_async_flush() nfs_write_error_remove_page(req);
0xffffffffc08b2614 <nfs_do_writepage+0x1b4>: mov %rbp,%rdi # x86_64下整数参数使用的寄存器依次为: RDI，RSI，RDX，RCX，R8，R9
# fs/nfs/write.c: 664 nfs_page_async_flush() return 0;
0xffffffffc08b2617 <nfs_do_writepage+0x1b7>: xor %r14d,%r14d # 最终效果是将 %r14d 寄存器的值设置为零
# fs/nfs/write.c: 663 nfs_page_async_flush() nfs_write_error_remove_page(req);
0xffffffffc08b261a <nfs_do_writepage+0x1ba>: callq 0xffffffffc08b1580 <nfs_write_error_remove_page>
# fs/nfs/write.c: 679 return ret;
0xffffffffc08b261f <nfs_do_writepage+0x1bf>: mov %r14d,%eax # 返回值放在eax
0xffffffffc08b2622 <nfs_do_writepage+0x1c2>: pop %rbx
0xffffffffc08b2623 <nfs_do_writepage+0x1c3>: pop %rbp
0xffffffffc08b2624 <nfs_do_writepage+0x1c4>: pop %r12
0xffffffffc08b2626 <nfs_do_writepage+0x1c6>: pop %r13
0xffffffffc08b2628 <nfs_do_writepage+0x1c8>: pop %r14
0xffffffffc08b262a <nfs_do_writepage+0x1ca>: pop %r15
0xffffffffc08b262c <nfs_do_writepage+0x1cc>: retq
...
# fs/nfs/write.c: 675 redirty_page_for_writepage(wbc, page);
0xffffffffc08b2671 <nfs_do_writepage+0x211>: mov %rbx,%rsi
0xffffffffc08b2674 <nfs_do_writepage+0x214>: mov %r13,%rdi在sget_userns()中，如果指定不一样的挂载选项时（比如加了soft），会生成新的超级块；而如果挂载选项和其他挂载点一样，就会尝试获取已有的超级块，但其他挂载点对应的同一超级块的锁已经被其他进程持有，所以就出现hung住的情况:
mount
ksys_mount
do_mount
do_new_mount
vfs_kern_mount
mount_fs
nfs_fs_mount
nfs_try_mount
nfs_try_mount_request
nfs3_create_server
nfs_create_server
nfs_init_server
nfs_get_client
nfs_alloc_client
nfs_init_client
nfs_fs_mount_common
sget_userns
grab_super // 当挂载选项一样时，尝试获取已有的超级块
down_write // 其他挂载点对应的同一超级块的锁已经被其他进程持有
alloc_super // 只有挂载选项不同时，才会分配新的超级块
wb_workfn
wb_writeback
__writeback_inodes_wb
trylock_super(sb)
down_read_trylock(&sb->s_umount) // 持有超级块读锁
writeback_sb_inodes
__writeback_single_inode
do_writepages
nfs_writepages
write_cache_pages
nfs_writepages_callback
nfs_do_writepage
nfs_page_async_flush
nfs_write_error_remove_page
nfs_release_request
nfs_free_request
__put_nfs_open_context
dput
dentry_kill
__dentry_kill
evict
inode_wait_for_writeback
nfs_parse_mount_options
mnt->flags |= NFS_MOUNT_SOFT; // soft
mnt->acregmin = mnt->acregmax = mnt->acdirmin = mnt->acdirmax = option; // actimeo#include <fcntl.h>
#include <unistd.h>
#include <sys/stat.h>
#include <string.h>
int main() {
const char *filename = "/mnt/file"; // NFS文件路径
const char *data = "hello\n"; // 要写入的数据
// 打开文件（如果不存在则创建，只写模式）
int fd = open(filename, O_CREAT | O_WRONLY, 0644);
if (fd == -1) {
// 错误处理：无法打开文件
return 1;
}
// 计算数据长度并写入文件
ssize_t bytes_written = write(fd, data, strlen(data));
if (bytes_written == -1) {
// 错误处理：写入失败
close(fd);
return 1;
}
// 无限循环，保持程序运行且不关闭文件
while (1) {
sleep(1); // 每秒休眠一次，减少CPU占用
}
// 关闭文件
close(fd);
return 0;
}默认选项值acregmin=3,acregmax=60,acdirmin=30,acdirmax=60，只需要更改其中一个选项，如acdirmax=59就能在挂载时生成新的超级块，从而挂载成功。