| Message ID | 20240717124534.1200735-1-cleger@rivosinc.com |
|---|---|
| State | New |
| Headers | show |
| Series | [v4] osdep: add a qemu_close_all_open_fd() helper | expand |
Hi Clément, On 17/7/24 14:45, Clément Léger wrote: > Since commit 03e471c41d8b ("qemu_init: increase NOFILE soft limit on > POSIX"), the maximum number of file descriptors that can be opened are > raised to nofile.rlim_max. On recent debian distro, this yield a maximum > of 1073741816 file descriptors. Now, when forking to start > qemu-bridge-helper, this actually calls close() on the full possible file > descriptor range (more precisely [3 - sysconf(_SC_OPEN_MAX)]) which > takes a considerable amount of time. In order to reduce that time, > factorize existing code to close all open files descriptors in a new > qemu_close_all_open_fd() function. This function uses various methods > to close all the open file descriptors ranging from the most efficient > one to the least one. It also accepts an ordered array of file > descriptors that should not be closed since this is required by the > callers that calls it after forking. > > Signed-off-by: Clément Léger <cleger@rivosinc.com> > Reviewed-by: Richard Henderson <richard.henderson@linaro.org> > ---- FYI git tools parse 3 '-', not 4. > v4: > - Add a comment saying that qemu_close_all_fds() can take a NULL skip > array and nskip == 0 > - Added an assert in qemu_close_all_fds() to check for skip/nskip > parameters > - Fix spurious tabs instead of spaces > - Applied checkpatch > - v3: https://lore.kernel.org/qemu-devel/20240716144006.6571-1-cleger@rivosinc.com/ > +void qemu_close_all_open_fd(const int *skip, unsigned int nskip) > +{ > + int open_max = sysconf(_SC_OPEN_MAX); > + unsigned int cur_skip = 0; > + int i; > + > + assert(skip != NULL || nskip == 0); > + > + if (qemu_close_all_open_fd_close_range(skip, nskip)) { > + return; > + } > + > + if (qemu_close_all_open_fd_proc(skip, nskip)) { > + return; > + } > + > + /* Fallback */ > + for (i = 0; i < open_max; i++) { > + if (cur_skip < nskip && i == skip[cur_skip]) { > + cur_skip++; > + continue; > + } > + close(i); > + } > +} Build failure on windows: ../util/osdep.c: In function 'qemu_close_all_open_fd': ../util/osdep.c:725:20: error: implicit declaration of function 'sysconf'; did you mean 'swscanf'? [-Wimplicit-function-declaration] 725 | int open_max = sysconf(_SC_OPEN_MAX); | ^~~~~~~ | swscanf ../util/osdep.c:725:20: error: nested extern declaration of 'sysconf' [-Werror=nested-externs] ../util/osdep.c:725:28: error: '_SC_OPEN_MAX' undeclared (first use in this function); did you mean 'FOPEN_MAX'? 725 | int open_max = sysconf(_SC_OPEN_MAX); | ^~~~~~~~~~~~ | FOPEN_MAX ../util/osdep.c:725:28: note: each undeclared identifier is reported only once for each function it appears in
On 23/07/2024 08:24, Philippe Mathieu-Daudé wrote: > Hi Clément, > > On 17/7/24 14:45, Clément Léger wrote: >> Since commit 03e471c41d8b ("qemu_init: increase NOFILE soft limit on >> POSIX"), the maximum number of file descriptors that can be opened are >> raised to nofile.rlim_max. On recent debian distro, this yield a maximum >> of 1073741816 file descriptors. Now, when forking to start >> qemu-bridge-helper, this actually calls close() on the full possible file >> descriptor range (more precisely [3 - sysconf(_SC_OPEN_MAX)]) which >> takes a considerable amount of time. In order to reduce that time, >> factorize existing code to close all open files descriptors in a new >> qemu_close_all_open_fd() function. This function uses various methods >> to close all the open file descriptors ranging from the most efficient >> one to the least one. It also accepts an ordered array of file >> descriptors that should not be closed since this is required by the >> callers that calls it after forking. >> >> Signed-off-by: Clément Léger <cleger@rivosinc.com> >> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> >> ---- > > FYI git tools parse 3 '-', not 4. Hi Philippe, Thanks for the info, I was not aware of that ! I'll use 3 '-' from now on. > >> v4: >> - Add a comment saying that qemu_close_all_fds() can take a NULL skip >> array and nskip == 0 >> - Added an assert in qemu_close_all_fds() to check for skip/nskip >> parameters >> - Fix spurious tabs instead of spaces >> - Applied checkpatch >> - v3: >> https://lore.kernel.org/qemu-devel/20240716144006.6571-1-cleger@rivosinc.com/ > > >> +void qemu_close_all_open_fd(const int *skip, unsigned int nskip) >> +{ >> + int open_max = sysconf(_SC_OPEN_MAX); >> + unsigned int cur_skip = 0; >> + int i; >> + >> + assert(skip != NULL || nskip == 0); >> + >> + if (qemu_close_all_open_fd_close_range(skip, nskip)) { >> + return; >> + } >> + >> + if (qemu_close_all_open_fd_proc(skip, nskip)) { >> + return; >> + } >> + >> + /* Fallback */ >> + for (i = 0; i < open_max; i++) { >> + if (cur_skip < nskip && i == skip[cur_skip]) { >> + cur_skip++; >> + continue; >> + } >> + close(i); >> + } >> +} > > Build failure on windows: > > ../util/osdep.c: In function 'qemu_close_all_open_fd': > ../util/osdep.c:725:20: error: implicit declaration of function > 'sysconf'; did you mean 'swscanf'? [-Wimplicit-function-declaration] > 725 | int open_max = sysconf(_SC_OPEN_MAX); > | ^~~~~~~ > | swscanf > ../util/osdep.c:725:20: error: nested extern declaration of 'sysconf' > [-Werror=nested-externs] > ../util/osdep.c:725:28: error: '_SC_OPEN_MAX' undeclared (first use in > this function); did you mean 'FOPEN_MAX'? > 725 | int open_max = sysconf(_SC_OPEN_MAX); > | ^~~~~~~~~~~~ > | FOPEN_MAX > ../util/osdep.c:725:28: note: each undeclared identifier is reported > only once for each function it appears in > Should I move this chunk of code in oslib-posix.c and stub that function for win32 ? It seems like this code was not built for windows before I added it in osdep, which means it is not needed for win32. What's your advice on that ? Thanks, Clément
Hi On Wed, Jul 17, 2024 at 4:48 PM Clément Léger <cleger@rivosinc.com> wrote: > Since commit 03e471c41d8b ("qemu_init: increase NOFILE soft limit on > POSIX"), the maximum number of file descriptors that can be opened are > raised to nofile.rlim_max. On recent debian distro, this yield a maximum > of 1073741816 file descriptors. Now, when forking to start > qemu-bridge-helper, this actually calls close() on the full possible file > descriptor range (more precisely [3 - sysconf(_SC_OPEN_MAX)]) which > takes a considerable amount of time. In order to reduce that time, > factorize existing code to close all open files descriptors in a new > qemu_close_all_open_fd() function. This function uses various methods > to close all the open file descriptors ranging from the most efficient > one to the least one. It also accepts an ordered array of file > descriptors that should not be closed since this is required by the > callers that calls it after forking. > > Signed-off-by: Clément Léger <cleger@rivosinc.com> > Reviewed-by: Richard Henderson <richard.henderson@linaro.org> > GLib already implemented those kinds of portable facilities. I wonder why launch_script() is not using glib gspawn API. async-teardown should use g_clownfrom() when glib >= 2.80. my 2c
On Tue, Jul 23, 2024 at 09:16:15AM +0200, Clément Léger wrote: > > > On 23/07/2024 08:24, Philippe Mathieu-Daudé wrote: > > Hi Clément, > > > > On 17/7/24 14:45, Clément Léger wrote: > >> Since commit 03e471c41d8b ("qemu_init: increase NOFILE soft limit on > >> POSIX"), the maximum number of file descriptors that can be opened are > >> raised to nofile.rlim_max. On recent debian distro, this yield a maximum > >> of 1073741816 file descriptors. Now, when forking to start > >> qemu-bridge-helper, this actually calls close() on the full possible file > >> descriptor range (more precisely [3 - sysconf(_SC_OPEN_MAX)]) which > >> takes a considerable amount of time. In order to reduce that time, > >> factorize existing code to close all open files descriptors in a new > >> qemu_close_all_open_fd() function. This function uses various methods > >> to close all the open file descriptors ranging from the most efficient > >> one to the least one. It also accepts an ordered array of file > >> descriptors that should not be closed since this is required by the > >> callers that calls it after forking. > >> > >> Signed-off-by: Clément Léger <cleger@rivosinc.com> > >> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> > >> ---- > > > > FYI git tools parse 3 '-', not 4. > > Hi Philippe, > > Thanks for the info, I was not aware of that ! I'll use 3 '-' from now on. > > > > >> v4: > >> - Add a comment saying that qemu_close_all_fds() can take a NULL skip > >> array and nskip == 0 > >> - Added an assert in qemu_close_all_fds() to check for skip/nskip > >> parameters > >> - Fix spurious tabs instead of spaces > >> - Applied checkpatch > >> - v3: > >> https://lore.kernel.org/qemu-devel/20240716144006.6571-1-cleger@rivosinc.com/ > > > > > >> +void qemu_close_all_open_fd(const int *skip, unsigned int nskip) > >> +{ > >> + int open_max = sysconf(_SC_OPEN_MAX); > >> + unsigned int cur_skip = 0; > >> + int i; > >> + > >> + assert(skip != NULL || nskip == 0); > >> + > >> + if (qemu_close_all_open_fd_close_range(skip, nskip)) { > >> + return; > >> + } > >> + > >> + if (qemu_close_all_open_fd_proc(skip, nskip)) { > >> + return; > >> + } > >> + > >> + /* Fallback */ > >> + for (i = 0; i < open_max; i++) { > >> + if (cur_skip < nskip && i == skip[cur_skip]) { > >> + cur_skip++; > >> + continue; > >> + } > >> + close(i); > >> + } > >> +} > > > > Build failure on windows: > > > > ../util/osdep.c: In function 'qemu_close_all_open_fd': > > ../util/osdep.c:725:20: error: implicit declaration of function > > 'sysconf'; did you mean 'swscanf'? [-Wimplicit-function-declaration] > > 725 | int open_max = sysconf(_SC_OPEN_MAX); > > | ^~~~~~~ > > | swscanf > > ../util/osdep.c:725:20: error: nested extern declaration of 'sysconf' > > [-Werror=nested-externs] > > ../util/osdep.c:725:28: error: '_SC_OPEN_MAX' undeclared (first use in > > this function); did you mean 'FOPEN_MAX'? > > 725 | int open_max = sysconf(_SC_OPEN_MAX); > > | ^~~~~~~~~~~~ > > | FOPEN_MAX > > ../util/osdep.c:725:28: note: each undeclared identifier is reported > > only once for each function it appears in > > > > Should I move this chunk of code in oslib-posix.c and stub that function > for win32 ? It seems like this code was not built for windows before I > added it in osdep, which means it is not needed for win32. Yes, I think that'll be OK. The fork/exec concept isn't present on Windows so we also don't need to close all FDs. With regards, Daniel
diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h index 191916f38e..43a035d756 100644 --- a/include/qemu/osdep.h +++ b/include/qemu/osdep.h @@ -757,6 +757,14 @@ static inline void qemu_reset_optind(void) int qemu_fdatasync(int fd); +/** + * Close all open file descriptors except the ones supplied in the @skip array + * + * @skip: ordered array of distinct file descriptors that should not be closed + * @nskip: number of entries in the @skip array. + */ +void qemu_close_all_open_fd(const int *skip, unsigned int nskip); + /** * Sync changes made to the memory mapped file back to the backing * storage. For POSIX compliant systems this will fallback diff --git a/net/tap.c b/net/tap.c index 51f7aec39d..45cb963142 100644 --- a/net/tap.c +++ b/net/tap.c @@ -385,6 +385,23 @@ static TAPState *net_tap_fd_init(NetClientState *peer, return s; } +static void close_all_fds_after_fork(int excluded_fd) +{ + const int skip_fd[] = {STDIN_FILENO, STDOUT_FILENO, STDERR_FILENO, + excluded_fd}; + unsigned int nskip = ARRAY_SIZE(skip_fd); + + /* + * skip_fd must be an ordered array of distinct fds, exclude + * excluded_fd if already included in the [STDIN_FILENO - STDERR_FILENO] + * range + */ + if (excluded_fd <= STDERR_FILENO) { + nskip--; + } + qemu_close_all_open_fd(skip_fd, nskip); +} + static void launch_script(const char *setup_script, const char *ifname, int fd, Error **errp) { @@ -400,13 +417,7 @@ static void launch_script(const char *setup_script, const char *ifname, return; } if (pid == 0) { - int open_max = sysconf(_SC_OPEN_MAX), i; - - for (i = 3; i < open_max; i++) { - if (i != fd) { - close(i); - } - } + close_all_fds_after_fork(fd); parg = args; *parg++ = (char *)setup_script; *parg++ = (char *)ifname; @@ -490,17 +501,11 @@ static int net_bridge_run_helper(const char *helper, const char *bridge, return -1; } if (pid == 0) { - int open_max = sysconf(_SC_OPEN_MAX), i; char *fd_buf = NULL; char *br_buf = NULL; char *helper_cmd = NULL; - for (i = 3; i < open_max; i++) { - if (i != sv[1]) { - close(i); - } - } - + close_all_fds_after_fork(sv[1]); fd_buf = g_strdup_printf("%s%d", "--fd=", sv[1]); if (strrchr(helper, ' ') || strrchr(helper, '\t')) { diff --git a/system/async-teardown.c b/system/async-teardown.c index 396963c091..9148ee8d04 100644 --- a/system/async-teardown.c +++ b/system/async-teardown.c @@ -26,40 +26,6 @@ static pid_t the_ppid; -/* - * Close all open file descriptors. - */ -static void close_all_open_fd(void) -{ - struct dirent *de; - int fd, dfd; - DIR *dir; - -#ifdef CONFIG_CLOSE_RANGE - int r = close_range(0, ~0U, 0); - if (!r) { - /* Success, no need to try other ways. */ - return; - } -#endif - - dir = opendir("/proc/self/fd"); - if (!dir) { - /* If /proc is not mounted, there is nothing that can be done. */ - return; - } - /* Avoid closing the directory. */ - dfd = dirfd(dir); - - for (de = readdir(dir); de; de = readdir(dir)) { - fd = atoi(de->d_name); - if (fd != dfd) { - close(fd); - } - } - closedir(dir); -} - static void hup_handler(int signal) { /* Check every second if this process has been reparented. */ @@ -85,9 +51,8 @@ static int async_teardown_fn(void *arg) /* * Close all file descriptors that might have been inherited from the * main qemu process when doing clone, needed to make libvirt happy. - * Not using close_range for increased compatibility with older kernels. */ - close_all_open_fd(); + qemu_close_all_open_fd(NULL, 0); /* Set up a handler for SIGHUP and unblock SIGHUP. */ sigaction(SIGHUP, &sa, NULL); diff --git a/util/osdep.c b/util/osdep.c index 770369831b..2f95748a4a 100644 --- a/util/osdep.c +++ b/util/osdep.c @@ -613,3 +613,135 @@ int qemu_fdatasync(int fd) return fsync(fd); #endif } + +static bool qemu_close_all_open_fd_proc(const int *skip, unsigned int nskip) +{ +#ifdef CONFIG_LINUX + struct dirent *de; + int fd, dfd; + bool close_fd; + DIR *dir; + unsigned int i, skip_start = 0, skip_end = nskip; + + dir = opendir("/proc/self/fd"); + if (!dir) { + /* If /proc is not mounted, there is nothing that can be done. */ + return false; + } + /* Avoid closing the directory. */ + dfd = dirfd(dir); + + for (de = readdir(dir); de; de = readdir(dir)) { + if (de->d_name[0] == '.') { + continue; + } + fd = atoi(de->d_name); + close_fd = true; + if (fd == dfd) { + close_fd = false; + } else { + for (i = skip_start; i < skip_end; i++) { + if (fd < skip[i]) { + /* We are below the next skipped fd, break */ + break; + } else if (fd == skip[i]) { + close_fd = false; + /* Restrict the range as we found fds matching start/end */ + if (i == skip_start) { + skip_start++; + } else if (i == skip_end) { + skip_end--; + } + break; + } + } + } + if (close_fd) { + close(fd); + } + } + closedir(dir); + + return true; +#else + return false; +#endif +} + + +static bool qemu_close_all_open_fd_close_range(const int *skip, + unsigned int nskip) +{ +#ifdef CONFIG_CLOSE_RANGE + int max_fd = sysconf(_SC_OPEN_MAX) - 1; + int first = 0, last = max_fd; + unsigned int cur_skip = 0; + int ret; + + do { + /* Find the start boundary of the range to close */ + while (cur_skip < nskip && first == skip[cur_skip]) { + cur_skip++; + first++; + } + + /* Find the upper boundary of the range to close */ + if (cur_skip < nskip) { + last = skip[cur_skip] - 1; + } + /* + * Adjust the maximum fd to close if it's above what the system + * supports + */ + if (last > max_fd) { + last = max_fd; + /* + * We can directly skip the remaining skip fds since the current + * one is already above the maximum supported one. + */ + cur_skip = nskip; + } + /* If last was adjusted, we might be > first, bail out */ + if (first > last) { + break; + } + + ret = close_range(first, last, 0); + if (ret < 0) { + return false; + } + first = last + 1; + last = max_fd; + } while (cur_skip < nskip); + + return true; +#else + return false; +#endif +} + +void qemu_close_all_open_fd(const int *skip, unsigned int nskip) +{ + int open_max = sysconf(_SC_OPEN_MAX); + unsigned int cur_skip = 0; + int i; + + assert(skip != NULL || nskip == 0); + + if (qemu_close_all_open_fd_close_range(skip, nskip)) { + return; + } + + if (qemu_close_all_open_fd_proc(skip, nskip)) { + return; + } + + /* Fallback */ + for (i = 0; i < open_max; i++) { + if (cur_skip < nskip && i == skip[cur_skip]) { + cur_skip++; + continue; + } + close(i); + } +}
Since commit 03e471c41d8b ("qemu_init: increase NOFILE soft limit on POSIX"), the maximum number of file descriptors that can be opened are raised to nofile.rlim_max. On recent debian distro, this yield a maximum of 1073741816 file descriptors. Now, when forking to start qemu-bridge-helper, this actually calls close() on the full possible file descriptor range (more precisely [3 - sysconf(_SC_OPEN_MAX)]) which takes a considerable amount of time. In order to reduce that time, factorize existing code to close all open files descriptors in a new qemu_close_all_open_fd() function. This function uses various methods to close all the open file descriptors ranging from the most efficient one to the least one. It also accepts an ordered array of file descriptors that should not be closed since this is required by the callers that calls it after forking. Signed-off-by: Clément Léger <cleger@rivosinc.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> ---- v4: - Add a comment saying that qemu_close_all_fds() can take a NULL skip array and nskip == 0 - Added an assert in qemu_close_all_fds() to check for skip/nskip parameters - Fix spurious tabs instead of spaces - Applied checkpatch - v3: https://lore.kernel.org/qemu-devel/20240716144006.6571-1-cleger@rivosinc.com/ v3: - Use STD*_FILENO defines instead of raw values - Fix indentation of close_all_fds_after_fork() - Check for nksip in fallback code - Check for path starting with a '.' in qemu_close_all_open_fd_proc() - Use unsigned for cur_skip - Move ifdefs inside close_fds functions rather than redefining them - Remove uneeded 'if(nskip)' test - Add comments to close_range version - Reduce range of skip fd as we find them in - v2: https://lore.kernel.org/qemu-devel/20240618111704.63092-1-cleger@rivosinc.com/ v2: - Factorize async_teardown.c close_fds implementation as well as tap.c ones - Apply checkpatch - v1: https://lore.kernel.org/qemu-devel/20240617162520.4045016-1-cleger@rivosinc.com/ --- include/qemu/osdep.h | 8 +++ net/tap.c | 33 +++++----- system/async-teardown.c | 37 +---------- util/osdep.c | 132 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 160 insertions(+), 50 deletions(-)