I am running dbench benchmarking tool for fileio on my fuse filesystem implementation and I am making sure that that fuse can sustain the test.
I am use libfuse library (https://github.com/libfuse/libfuse) to implement fuse. I am not doing anything fancy, I am just made it a pass through, reads just do pread and writes just do pwrite.
I left it running for about 10 hours and I see a steady increase in memory from 0 to 21G. Has anyone else seen this? I checked for mem leaks in my code but I am doing really simple passthrough code that I no where allocate heap memory specifically. Like a simple read function looks like this. FUSE_DATA is a global struct that gets assigned in the main function.
void fuse_fullpath(char fpath[PATH_MAX], const char *path, const char * rootdir)
{
if (rootdir == NULL || path == NULL) {
return;
}
strcpy(fpath, rootdir);
strncat(fpath, path, PATH_MAX);
}
static int fuse_read(const char *path, char *buf, size_t size, off_t offset,
struct fuse_file_info *fi)
{
int res;
char fpath[PATH_MAX];
fuse_fullpath(fpath, path, FUSE_DATA->rootdir);
res = pread(fi->fh, buf, size, offset);
if (res == -1) {
res = -errno;
}
return res;
}
I tried valgrind on fuse and the report from it just gives me a lot in libgobject library like this.
==31923== 8 bytes in 1 blocks are still reachable in loss record 2 of 245
==31923== at 0x4C2CC70: calloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==31923== by 0x54E0668: g_malloc0 (in /lib/x86_64-linux-gnu/libglib-2.0.so.0.4002.0)
==31923== by 0x4E63315: ??? (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.4002.0)
==31923== by 0x4E67108: g_type_register_static (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.4002.0)
==31923== by 0x4E6FF71: g_pointer_type_register_static (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.4002.0)
==31923== by 0x4E6FFF7: g_gtype_get_type (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.4002.0)
==31923== by 0x4E558E7: ??? (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.4002.0)