| Message ID | 4d8628bf1d8630c16de30d47732058ec82293429.1579631655.git.fweimer@redhat.com |
|---|---|
| State | New |
| Headers | show |
| Series | Race condition in /etc/resolv.conf reloading (bug 25420) | expand |
On 21/01/2020 15:41, Florian Weimer wrote: > res_vinit_1 did not close the stream on errors, only on success. > This change moves closing the stream to __resolv_conf_load, for both > the success and error cases. > > Fixes commit 89f187a40fc0ad4e22838526bfe34d73f758b776 ("resolv: Use > getline for configuration file reading in res_vinit_1") and commit > 3f853f22c87f0b671c0366eb290919719fa56c0e ("resolv: Lift domain search > list limits [BZ #19569] [BZ #21475]"), where memory allocation was > introduced into res_vinit_1. > --- > resolv/res_init.c | 8 +++++++- > 1 file changed, 7 insertions(+), 1 deletion(-) > > diff --git a/resolv/res_init.c b/resolv/res_init.c > index 95dce098aa..09345718cd 100644 > --- a/resolv/res_init.c > +++ b/resolv/res_init.c > @@ -508,7 +508,6 @@ res_vinit_1 (FILE *fp, struct resolv_conf_parser *parser) > continue; > } > } > - fclose (fp); > } > if (__glibc_unlikely (nameserver_list_size (&parser->nameserver_list) == 0)) > { > @@ -593,6 +592,13 @@ __resolv_conf_load (struct __res_state *preinit) > } > resolv_conf_parser_free (&parser); > > + if (fp != NULL) > + { > + int saved_errno = errno; > + fclose (fp); > + __set_errno (saved_errno); > + } > + > return conf; > } Why not close the FILE on __resolv_conf_load? It make the FILE object cleanup as close as its creation, which usually improves readability.
* Adhemerval Zanella: > On 21/01/2020 15:41, Florian Weimer wrote: >> res_vinit_1 did not close the stream on errors, only on success. >> This change moves closing the stream to __resolv_conf_load, for both >> the success and error cases. >> >> Fixes commit 89f187a40fc0ad4e22838526bfe34d73f758b776 ("resolv: Use >> getline for configuration file reading in res_vinit_1") and commit >> 3f853f22c87f0b671c0366eb290919719fa56c0e ("resolv: Lift domain search >> list limits [BZ #19569] [BZ #21475]"), where memory allocation was >> introduced into res_vinit_1. >> --- >> resolv/res_init.c | 8 +++++++- >> 1 file changed, 7 insertions(+), 1 deletion(-) >> >> diff --git a/resolv/res_init.c b/resolv/res_init.c >> index 95dce098aa..09345718cd 100644 >> --- a/resolv/res_init.c >> +++ b/resolv/res_init.c >> @@ -508,7 +508,6 @@ res_vinit_1 (FILE *fp, struct resolv_conf_parser *parser) >> continue; >> } >> } >> - fclose (fp); >> } >> if (__glibc_unlikely (nameserver_list_size (&parser->nameserver_list) == 0)) >> { >> @@ -593,6 +592,13 @@ __resolv_conf_load (struct __res_state *preinit) >> } >> resolv_conf_parser_free (&parser); >> >> + if (fp != NULL) >> + { >> + int saved_errno = errno; >> + fclose (fp); >> + __set_errno (saved_errno); >> + } >> + >> return conf; >> } > > Why not close the FILE on __resolv_conf_load? It make the FILE object > cleanup as close as its creation, which usually improves readability. Sorry, I don't understand. Isn't this what the patch does? Thanks, Florian
On 13/02/2020 18:08, Florian Weimer wrote: > * Adhemerval Zanella: > >> On 21/01/2020 15:41, Florian Weimer wrote: >>> res_vinit_1 did not close the stream on errors, only on success. >>> This change moves closing the stream to __resolv_conf_load, for both >>> the success and error cases. >>> >>> Fixes commit 89f187a40fc0ad4e22838526bfe34d73f758b776 ("resolv: Use >>> getline for configuration file reading in res_vinit_1") and commit >>> 3f853f22c87f0b671c0366eb290919719fa56c0e ("resolv: Lift domain search >>> list limits [BZ #19569] [BZ #21475]"), where memory allocation was >>> introduced into res_vinit_1. >>> --- >>> resolv/res_init.c | 8 +++++++- >>> 1 file changed, 7 insertions(+), 1 deletion(-) >>> >>> diff --git a/resolv/res_init.c b/resolv/res_init.c >>> index 95dce098aa..09345718cd 100644 >>> --- a/resolv/res_init.c >>> +++ b/resolv/res_init.c >>> @@ -508,7 +508,6 @@ res_vinit_1 (FILE *fp, struct resolv_conf_parser *parser) >>> continue; >>> } >>> } >>> - fclose (fp); >>> } >>> if (__glibc_unlikely (nameserver_list_size (&parser->nameserver_list) == 0)) >>> { >>> @@ -593,6 +592,13 @@ __resolv_conf_load (struct __res_state *preinit) >>> } >>> resolv_conf_parser_free (&parser); >>> >>> + if (fp != NULL) >>> + { >>> + int saved_errno = errno; >>> + fclose (fp); >>> + __set_errno (saved_errno); >>> + } >>> + >>> return conf; >>> } >> >> Why not close the FILE on __resolv_conf_load? It make the FILE object >> cleanup as close as its creation, which usually improves readability. > > Sorry, I don't understand. Isn't this what the patch does? Nevermind, for some reason I though it was doing on res_vinit_1. LGTM, thanks. Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
diff --git a/resolv/res_init.c b/resolv/res_init.c index 95dce098aa..09345718cd 100644 --- a/resolv/res_init.c +++ b/resolv/res_init.c @@ -508,7 +508,6 @@ res_vinit_1 (FILE *fp, struct resolv_conf_parser *parser) continue; } } - fclose (fp); } if (__glibc_unlikely (nameserver_list_size (&parser->nameserver_list) == 0)) { @@ -593,6 +592,13 @@ __resolv_conf_load (struct __res_state *preinit) } resolv_conf_parser_free (&parser); + if (fp != NULL) + { + int saved_errno = errno; + fclose (fp); + __set_errno (saved_errno); + } + return conf; }
res_vinit_1 did not close the stream on errors, only on success. This change moves closing the stream to __resolv_conf_load, for both the success and error cases. Fixes commit 89f187a40fc0ad4e22838526bfe34d73f758b776 ("resolv: Use getline for configuration file reading in res_vinit_1") and commit 3f853f22c87f0b671c0366eb290919719fa56c0e ("resolv: Lift domain search list limits [BZ #19569] [BZ #21475]"), where memory allocation was introduced into res_vinit_1. --- resolv/res_init.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)