lix-project/lix
20
58
Fork 31
Code Issues 364 Code Review (Gerrit) Projects 3 Releases Wiki Activity

libstore/linux: precompile and cache the seccomp BPF

Browse source 
The growth of the seccomp filter in 127ee1a101
made its compilation time significant (roughly 10 milliseconds have been
measured on one machine). For this reason, it is now precompiled and cached in
the parent process so that this overhead is not hit for every single build. It
is still not optimal when going through the daemon, because compilation still
happens once per client, but it's better than before and doing it only once for
the entire daemon requires excessive crimes with the current architecture.

Fixes: #461
Change-Id: I2277eaaf6bab9bd74bbbfd9861e52392a54b61a3
This commit is contained in:
alois31 2024-08-06 16:38:32 +02:00
parent 403fa9e2b6
commit 2c48460850
Signed by: alois31
GPG key ID: E0F59EA5E5216914
1 changed files with 16 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
src/libstore/platform/linux.cc
View file

@ -717,6 +717,12 @@ static std::vector<struct sock_filter> compileSyscallFilter()
return filter;
}
static const std::vector<struct sock_filter> &getSyscallFilter()
{
static auto filter = compileSyscallFilter();
return filter;
}
#endif
void LinuxLocalDerivationGoal::setupSyscallFilter()
@ -727,11 +733,12 @@ void LinuxLocalDerivationGoal::setupSyscallFilter()
if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) == -1)
throw SysError("PR_SET_NO_NEW_PRIVS failed");
#if HAVE_SECCOMP
auto seccompBPF = compileSyscallFilter();
const auto &seccompBPF = getSyscallFilter();
assert(seccompBPF.size() <= std::numeric_limits<unsigned short>::max());
struct sock_fprog fprog = {
.len = static_cast<unsigned short>(seccompBPF.size()),
.filter = seccompBPF.data(),
// the kernel does not actually write to the filter
.filter = const_cast<struct sock_filter *>(seccompBPF.data()),
};
if (syscall(SYS_seccomp, SECCOMP_SET_MODE_FILTER, 0, &fprog) != 0)
throw SysError("unable to load seccomp BPF program");
@ -827,6 +834,13 @@ void LinuxLocalDerivationGoal::prepareSandbox()
Pid LinuxLocalDerivationGoal::startChild(std::function<void()> openSlave)
{
#if HAVE_SECCOMP
// Our seccomp filter program is surprisingly expensive to compile (~10ms).
// For this reason, we precompile it once and then cache it.
// This has to be done in the parent so that all builds get to use the same cache.
getSyscallFilter();
#endif
// If we're not sandboxing no need to faff about, use the fallback
if (!useChroot) {
return LocalDerivationGoal::startChild(openSlave);