|SUCCESS| pw159212 [PATCH v5 01/17] eal: use C library to parse filesystem table

[0-day Robot <robot@bytheb.org>]

Test-Label: 0day Robot - Claude SONNET
Test-Status: SUCCESS
http://dpdk.org/patch/159212

_AI Review Comments Available_


Git SHA: cd60dcd503
Author: Stephen Hemminger <stephen@networkplumber.org>
Subject: eal: use C library to parse filesystem table

This patch replaces custom string parsing of /proc/mounts with standard
C library routines getmntent_r() and related functions to parse the
filesystem mount table.

> Rather than doing parsing of /proc/mounts with open coded string
> handling, use the standard C library routines.
> These exist in BSD and Linux.
> 
> It also avoids any possible issues with escaped strings etc
> which the library handles. See getmntent(3) man page.
> 
> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
>  lib/eal/linux/eal_hugepage_info.c | 58 ++++++++++---------------------
>  1 file changed, 18 insertions(+), 40 deletions(-)
> 
> diff --git a/lib/eal/linux/eal_hugepage_info.c b/lib/eal/linux/eal_hugepage_info.c
> index d47a19c56a..7161b1a2fb 100644
> --- a/lib/eal/linux/eal_hugepage_info.c
> +++ b/lib/eal/linux/eal_hugepage_info.c
> @@ -13,6 +13,7 @@
>  #include <inttypes.h>
>  #include <unistd.h>
>  #include <errno.h>
> +#include <mntent.h>
>  #include <sys/mman.h>
>  #include <sys/stat.h>

[ ... ]

> @@ -226,56 +217,43 @@ get_hugepage_dir(uint64_t hugepage_sz, char *hugedir, int len)
>  		return -1;
>  	}
>  
> -	FILE *fd = fopen(proc_mounts, "r");
> -	if (fd == NULL)
> +	FILE *mounts = setmntent(proc_mounts, "r");
> +	if (mounts == NULL)
>  		rte_panic("Cannot open %s\n", proc_mounts);
>  
>  	if (default_size == 0)
>  		default_size = get_default_hp_size();
>  
> -	while (fgets(buf, sizeof(buf), fd)){
> +	while (getmntent_r(mounts, &entry, buf, sizeof(buf))) {

Does this code check for ENOENT or ERANGE return values from 
getmntent_r()? The function can return NULL but set errno to indicate
different types of failures, and partial reads might cause issues.

>  		const char *pagesz_str;
> -		size_t mountpt_len = 0;
> -
> -		if (rte_strsplit(buf, sizeof(buf), splitstr, _FIELDNAME_MAX,
> -				split_tok) != _FIELDNAME_MAX) {
> -			EAL_LOG(ERR, "Error parsing %s", proc_mounts);
> -			break; /* return NULL */
> -		}
> +		uint64_t pagesz = default_size;
> +		size_t mountpt_len;
>  
> -		if (strncmp(splitstr[FSTYPE], hugetlbfs_str, htlbfs_str_len) != 0)
> +		if (strcmp(entry.mnt_type, "hugetlbfs") != 0)
>  			continue;
>  
> -		pagesz_str = strstr(splitstr[OPTIONS], pagesize_opt);
> +		pagesz_str = strstr(entry.mnt_opts, pagesize_opt);
> +		if (pagesz_str)
> +			pagesz = rte_str_to_size(&pagesz_str[pagesize_opt_len]);

Could entry.mnt_opts be NULL? The mntent structure fields might not 
always be populated, and strstr() would segfault on a NULL pointer.

Also, does rte_str_to_size() handle malformed input gracefully? If the
pagesize option exists but has invalid data, this could cause issues.

[ ... ]