From: Xiaoming Ni <nixiaoming@huawei.com>
To: Adhemerval Zanella <adhemerval.zanella@linaro.org>,
<libc-alpha@sourceware.org>, <glibc-bugs@sourceware.org>,
<unassigned@sourceware.org>, <drepper.fsp@gmail.com>,
<roland@gnu.org>, <carlos@redhat.com>
Cc: <wangle6@huawei.com>, <yukeji@huawei.com>,
Paul Eggert <eggert@cs.ucla.edu>
Subject: Re: [PATCH v2] stdlib: realpath use malloc replace __alloca to reduce stack overflow risks [BZ #26341]
Date: Sat, 8 Aug 2020 17:14:34 +0800 [thread overview]
Message-ID: <a92abe8a-f1c4-71c3-c0c0-dff823f66435@huawei.com> (raw)
In-Reply-To: <ad170aab-3c41-af17-7320-7f68bf464462@linaro.org>
On 2020/8/8 3:43, Adhemerval Zanella wrote:
>
>
> On 07/08/2020 07:16, Xiaoming Ni wrote:
>> Realpath() cyclically invokes __alloca() when processing soft link files,
>> which may consume 164 KB stack space.
>> Therefore, replace __alloca with malloc to reduce stack overflow risks
>>
>> v2: Avoid repeated malloc and free operations. and add testcase
>> v1: https://patches-gcc.linaro.org/patch/39851/
>>
>> Signed-off-by: Xiaoming Ni <nixiaoming@huawei.com>
>
> Again, we do not use DCO, but rather Copyright assignment.
Do I just need to delete this line: "Signed-off-by: Xiaoming Ni
<nixiaoming@huawei.com>"?
Does this mean that glibc doesn't use
Signed-off-by/Reported-by/Tested-by/Reviewed-by?
> Paul, it might be something to be fixed on gnulib since I noted that both
> gpl and lgpl code uses malloca (which calls alloca if the header is present).
>
> Some comments below regarding testing.
>
>> ---
>> stdlib/Makefile | 3 +-
>> stdlib/canonicalize.c | 25 +++++++++++++--
>> stdlib/tst-bz26341.c | 73 +++++++++++++++++++++++++++++++++++++++++++
>> 3 files changed, 98 insertions(+), 3 deletions(-)
>> create mode 100644 stdlib/tst-bz26341.c
>>
>> diff --git a/stdlib/Makefile b/stdlib/Makefile
>> index 4615f6dfe7..bfdd9036b1 100644
>> --- a/stdlib/Makefile
>> +++ b/stdlib/Makefile
>> @@ -87,7 +87,7 @@ tests := tst-strtol tst-strtod testmb testrand testsort testdiv \
>> tst-makecontext-align test-bz22786 tst-strtod-nan-sign \
>> tst-swapcontext1 tst-setcontext4 tst-setcontext5 \
>> tst-setcontext6 tst-setcontext7 tst-setcontext8 \
>> - tst-setcontext9 tst-bz20544
>> + tst-setcontext9 tst-bz20544 tst-bz26341
>>
>> tests-internal := tst-strtod1i tst-strtod3 tst-strtod4 tst-strtod5i \
>> tst-tls-atexit tst-tls-atexit-nodelete
>
> Ok.
>
>> @@ -98,6 +98,7 @@ ifeq ($(build-hardcoded-path-in-tests),yes)
>> tests += tst-empty-env
>> endif
>>
>> +LDLIBS-tst-bz26341 = -lpthread
>
> This needs to be $(shared-thread-library).
>
>> LDLIBS-test-atexit-race = $(shared-thread-library)
>> LDLIBS-test-at_quick_exit-race = $(shared-thread-library)
>> LDLIBS-test-cxa_atexit-race = $(shared-thread-library)
>> diff --git a/stdlib/canonicalize.c b/stdlib/canonicalize.c
>> index cbd885a3c5..c02a8a5800 100644
>> --- a/stdlib/canonicalize.c
>> +++ b/stdlib/canonicalize.c
>> @@ -46,6 +46,7 @@ __realpath (const char *name, char *resolved)
>> const char *start, *end, *rpath_limit;
>> long int path_max;
>> int num_links = 0;
>> + char *buf = NULL;
>>
>> if (name == NULL)
>> {
>
> Ok.
>
>> @@ -163,9 +164,18 @@ __realpath (const char *name, char *resolved)
>>
>> if (S_ISLNK (st.st_mode))
>> {
>> - char *buf = __alloca (path_max);
>> size_t len;
>>
>> + if (buf == NULL)
>> + {
>> + buf = malloc (path_max);
>> + if (buf == NULL)
>> + {
>> + __set_errno (ENOMEM);
>> + goto error;
>> + }
>> + }
>> +
>> if (++num_links > __eloop_threshold ())
>> {
>> __set_errno (ELOOP);
>
> Ok.
>
>> @@ -178,7 +188,14 @@ __realpath (const char *name, char *resolved)
>> buf[n] = '\0';
>>
>> if (!extra_buf)
>> - extra_buf = __alloca (path_max);
>> + {
>> + extra_buf = malloc (path_max);
>> + if (extra_buf == NULL)
>> + {
>> + __set_errno (ENOMEM);
>> + goto error;
>> + }
>> + }
>>
>> len = strlen (end);
>> if (path_max - n <= len)
>
> Ok.
>
>> @@ -210,12 +227,16 @@ __realpath (const char *name, char *resolved)
>> *dest = '\0';
>>
>> assert (resolved == NULL || resolved == rpath);
>> + free(extra_buf);
>> + free(buf);
>> return rpath;
>>
>> error:
>> assert (resolved == NULL || resolved == rpath);
>> if (resolved == NULL)
>> free (rpath);
>> + free (extra_buf);
>> + free (buf);
>> return NULL;
>> }
>> libc_hidden_def (__realpath)
>
> Ok.
>
>> diff --git a/stdlib/tst-bz26341.c b/stdlib/tst-bz26341.c
>> new file mode 100644
>> index 0000000000..0fe095b7d1
>> --- /dev/null
>> +++ b/stdlib/tst-bz26341.c
>> @@ -0,0 +1,73 @@
>
> This test need the Copyright header and to be properly indented using glibc
> code guideline.
>
Is that it?
/* Copyright (C) 2020 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
> Use it need to use the libsupport (check support/README-testing.c and other
> tests that use '#include <support/test-driver.c>'.
sorry, I'm not familiar with the glibc test suite
If the test case is successful, 0 is returned. If the test case fails, a
non-zero value is returned. Is this sufficient?
I see that the main() function is also used in tst-qsort.c.
Is support/test-driver.c mandatory?
thanks
Xiaoming Ni
>
>
>> +#include <stdlib.h>
>> +#include <stdio.h>
>> +#include <unistd.h>
>> +#include <limits.h>
>> +#include <sys/resource.h>
>> +#include <sys/stat.h>
>> +#include <fcntl.h>
>> +#include <string.h>
>> +#include <pthread.h>
>> +
>> +int creat_link(void)
>> +{
>> + int i;
>> + int fd;
>> + char fname[2][16] = {0};
>> + char *p1 = fname[0];
>> + char *p2 = fname[1];
>> + strcpy(p1, "f0");
>> + fd = open(p1, O_RDONLY|O_CREAT, 0444);
>> + close(fd);
>> +
>> + for (i = 0; i < 41; i++) {
>> + sprintf(p2, "f%d", i);
>> + symlink(p1, p2);
>> + p1 = p2;
>> + p2 = fname[i % 2];
>> + }
>> + return 0;
>> +}
>> +
>> +void clean_link(void)
>> +{
>> + char fname[16] = {0};
>> + int i;
>> + for (i = 0; i < 41; i++) {
>> + sprintf(fname, "f%d", i);
>> + unlink(fname);
>> + }
>> +}
>> +
>> +void *do_realpath(void *ignore)
>> +{
>> + char *p = realpath("f40", NULL);
>> + printf("%p\n", p);
>> + if (p != NULL)
>> + printf("%s\n", p);
>> + return NULL;
>> +}
>> +
>> +/* Set different stack sizes and check whether stack overflow occurs. */
>> +int do_test(int size)
>> +{
>> + pthread_t tid;
>> + pthread_attr_t thread_attr;
>> + pthread_attr_init(&thread_attr);
>> + pthread_attr_setstacksize(&thread_attr, size);
>> +
>> + pthread_create(&tid, &thread_attr, do_realpath, NULL);
>> + pthread_join(tid, NULL);
>> + return 0;
>> +}
>> +
>> +int main(int argc, char *argv[])
>> +{
>> + creat_link();
>> + do_test(8192*1024);
>> + do_test(160*1024);
>> + do_test(40*1024);
>
> It think it would be suffice to just check with a small stacksize that triggers
> the failure on current code.
Yes, you're right.
>
>> + clean_link();
>> + printf("\n");
>> + return 0;
>> +}
>> +
>>
>
> .
>
Thanks
Xiaoming Ni
next prev parent reply other threads:[~2020-08-08 9:14 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-07 10:16 Xiaoming Ni
2020-08-07 19:43 ` Adhemerval Zanella
2020-08-08 0:00 ` Paul Eggert
2020-08-08 9:14 ` Xiaoming Ni [this message]
2020-08-07 23:56 ` Paul Eggert
2020-08-08 8:54 ` Xiaoming Ni
2020-08-09 8:44 ` Paul Eggert
2020-08-09 12:38 ` Florian Weimer
2020-08-09 17:22 ` Paul Eggert
2020-08-10 13:40 ` Adhemerval Zanella
2020-08-11 9:54 ` Paul Eggert
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=a92abe8a-f1c4-71c3-c0c0-dff823f66435@huawei.com \
--to=nixiaoming@huawei.com \
--cc=adhemerval.zanella@linaro.org \
--cc=carlos@redhat.com \
--cc=drepper.fsp@gmail.com \
--cc=eggert@cs.ucla.edu \
--cc=glibc-bugs@sourceware.org \
--cc=libc-alpha@sourceware.org \
--cc=roland@gnu.org \
--cc=unassigned@sourceware.org \
--cc=wangle6@huawei.com \
--cc=yukeji@huawei.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).