Fixes:
02021-10-18T09:12:31Z [init] [httpserver.cpp:181] [InitHTTPAllowList] Allowing HTTP connections from: 127.0.0.0/8 ::1/128
12021-10-18T09:12:31Z [init] [util/syscall_sandbox.cpp:487] [SyscallSandboxDebugSignalHandler] ERROR: The syscall "gettimeofday" (syscall number 96) is not allowed by the syscall sandbox in thread "init". Please report.
What environment/configuration is this? Interesting that we haven’t encountered this earlier :)
I guess one reason we haven’t seen this one earlier is that the library function gettimeofday can be implemented using the syscall clock_gettime (instead of the actual gettimeofday syscall).
176@@ -177,6 +177,7 @@ const std::map<uint32_t, std::string> LINUX_SYSCALLS{
177 {__NR_geteuid, "geteuid"},
178 {__NR_getgid, "getgid"},
179 {__NR_getgroups, "getgroups"},
180+ {__NR_gettimeofday, "gettimeofday"},
__NR_gettimeofday is already included in LINUX_SYSCALLS below (after gettid).
gettimeofday. This is much faster as the time is in a shared mapped page. However, if this isn’t available for some reason, libc take the slow path by calling the actual system call.
More context:
0 Why does the vDSO exist at all? There are some system calls the
1 kernel provides that user-space code ends up using frequently, to
2 the point that such calls can dominate overall performance. This
3 is due both to the frequency of the call as well as the context-
4 switch overhead that results from exiting user space and entering
5 the kernel.
6...
7 One frequently used system call is gettimeofday(2). This system
8 call is called both directly by user-space applications as well
9 as indirectly by the C library. Think timestamps or timing loops
10 or polling—all of these frequently need to know what time it is
11 right now. This information is also not secret—any application
12 in any privilege mode (root or any unprivileged user) will get
13 the same answer. Thus the kernel arranges for the information
14 required to answer this question to be placed in memory the
15 process can access. Now a call to gettimeofday(2) changes from a
16 system call to a normal function call and a few memory accesses.
17...
18 x86-64 functions
19 The table below lists the symbols exported by the vDSO. All of
20 these symbols are also available without the "__vdso_" prefix,
21 but you should ignore those and stick to the names below.
22
23 symbol version
24 ─────────────────────────────────
25 __vdso_clock_gettime LINUX_2.6
26 __vdso_getcpu LINUX_2.6
27 __vdso_gettimeofday LINUX_2.6
28 __vdso_time LINUX_2.6
Concept ACK (obviously)
And cr ACK once the removal of the duplicate LINUX_SYSCALLS is done :)
What environment/configuration is this? Interesting that we haven’t encountered this earlier :)
It might be required to change the timezone or so?
Also, sort entries. Can be reviewed with: --color-moved=dimmed-zebra
