I send v2 of this patch. I removed in v2 special buffer for file descriptor. It helped when we haved unitilized buffer[256] because of diferent fgets/fread/fwrite behaviour when the buffer was set.
The real and only problem was uninitialized buffer[256] and this workaround IO buffer is not needed for patch. > -----Original Message----- > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Daniel Mrzyglod > Sent: Friday, June 26, 2015 10:37 AM > To: dev at dpdk.org > Subject: [dpdk-dev] [PATCH] cfgfile: fix unitialised buffer and improve > reading > from nfs filesystem. > > Nature of the problem was not initialised buffer[256] on special condition > there were probability that program will work on unitialised data that > could provide unexpected program behaviour. > > Adding additional transparent I/O buffer for I/O operations > improve reading on heavyloaded enviroments with NFS. > > Signed-off-by: Daniel Mrzyglod <danielx.t.mrzyglod at intel.com> > --- > lib/librte_cfgfile/rte_cfgfile.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/lib/librte_cfgfile/rte_cfgfile.c > b/lib/librte_cfgfile/rte_cfgfile.c > index b81c273..88fcb46 100644 > --- a/lib/librte_cfgfile/rte_cfgfile.c > +++ b/lib/librte_cfgfile/rte_cfgfile.c > @@ -93,10 +93,14 @@ rte_cfgfile_load(const char *filename, int flags) > int curr_section = -1; > int curr_entry = -1; > char buffer[256]; > + char f_streambuff[BUFSIZ]; > int lineno = 0; > struct rte_cfgfile *cfg = NULL; > + memset(buffer, '\0', 256*sizeof(char)); > + memset(f_streambuff, '\0', BUFSIZ); > > FILE *f = fopen(filename, "r"); > + setbuf(f, f_streambuff); > if (f == NULL) > return NULL; > > -- > 2.1.0
