[BOLT] Avoid implicit memset in hugify

[yavtuk]

Summary

This patch removes memset usage from C library
relative to https://reviews.llvm.org/D128960 commit

[Amir]

Thanks for fixing it!

[Amir]

Before landing, please retitle the diff to e.g. "[BOLT] Avoid implicit memset in hugify", and amend the commit message to reflect the change.

[yavtuk]

updated commit message

[yavtuk]

@Amir Thanks for the review

[yota9]

Just curious is there any reason to memset this buffer? It will be overwritten by read anyway, probably buf[0] = '\0' is more then enough here. And looks like the size might be truncated to smth like 32-64, what do you think?

bolt/runtime/hugify.cpp

39

This is OOS but sizeof(buf) if you don't mind

[yavtuk]

In D128971#3624447, @yota9 wrote:

Just curious is there any reason to memset this buffer? It will be overwritten by read anyway, probably buf[0] = '\0' is more then enough here. And looks like the size might be truncated to smth like 32-64, what do you think?

I agree with __read, I am going to suggest a patch with Hugify support for PIE binaries soon, and this commit contains sizeof(buf).

buf[0] = '\0' vs memSet
In my thoughts both options are available here because it doesn’t affect performance

[maksfb]

Thanks for the fix.

[yavtuk]

fix in commit: https://reviews.llvm.org/D129168

bolt/runtime/hugify.cpp

 #ifdef MADV_HUGEPAGE
 /// Check whether the kernel supports THP via corresponding sysfs entry.
 static bool has_pagecache_thp_support() {
-  char buf[256] = {0};
+  char buf[256];
   const char *madviseStr = "always [madvise] never";
 
   int fd = __open("/sys/kernel/mm/transparent_hugepage/enabled",

   if (fd < 0)
     return false;
 
+  memSet(buf, 0, sizeof(buf));
   size_t res = __read(fd, buf, 256);
   if (res < 0)
     return false;