Date: Wed, 1 Aug 2018 19:06:55 +0000 (UTC) From: Konstantin Belousov <kib@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r337055 - head/sys/ufs/ffs Message-ID: <201808011906.w71J6tBe032796@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: kib Date: Wed Aug 1 19:06:55 2018 New Revision: 337055 URL: https://svnweb.freebsd.org/changeset/base/337055 Log: Avoid assertion in /dev/ufssuspend when the suspend ioctl is (incorrectly) called while another suspension is already active. PR: 230220 Reported by: dexuan Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 1 week Modified: head/sys/ufs/ffs/ffs_suspend.c Modified: head/sys/ufs/ffs/ffs_suspend.c ============================================================================== --- head/sys/ufs/ffs/ffs_suspend.c Wed Aug 1 19:02:05 2018 (r337054) +++ head/sys/ufs/ffs/ffs_suspend.c Wed Aug 1 19:06:55 2018 (r337055) @@ -218,6 +218,33 @@ ffs_susp_suspend(struct mount *mp) } static void +ffs_susp_unsuspend(struct mount *mp) +{ + struct ufsmount *ump; + + sx_assert(&ffs_susp_lock, SA_XLOCKED); + + /* + * XXX: The status is kept per-process; the vfs_write_resume() routine + * asserts that the resuming thread is the same one that called + * vfs_write_suspend(). The cdevpriv data, however, is attached + * to the file descriptor, e.g. is inherited during fork. Thus, + * it's possible that the resuming process will be different from + * the one that started the suspension. + * + * Work around by fooling the check in vfs_write_resume(). + */ + mp->mnt_susp_owner = curthread; + + vfs_write_resume(mp, 0); + ump = VFSTOUFS(mp); + UFS_LOCK(ump); + ump->um_flags &= ~UM_WRITESUSPENDED; + UFS_UNLOCK(ump); + vfs_unbusy(mp); +} + +static void ffs_susp_dtor(void *data) { struct fs *fs; @@ -243,24 +270,7 @@ ffs_susp_dtor(void *data) if (error != 0) panic("failed to unsuspend writes on %s", fs->fs_fsmnt); - /* - * XXX: The status is kept per-process; the vfs_write_resume() routine - * asserts that the resuming thread is the same one that called - * vfs_write_suspend(). The cdevpriv data, however, is attached - * to the file descriptor, e.g. is inherited during fork. Thus, - * it's possible that the resuming process will be different from - * the one that started the suspension. - * - * Work around by fooling the check in vfs_write_resume(). - */ - mp->mnt_susp_owner = curthread; - - vfs_write_resume(mp, 0); - vfs_unbusy(mp); - UFS_LOCK(ump); - ump->um_flags &= ~UM_WRITESUSPENDED; - UFS_UNLOCK(ump); - + ffs_susp_unsuspend(mp); sx_xunlock(&ffs_susp_lock); } @@ -300,7 +310,8 @@ ffs_susp_ioctl(struct cdev *dev, u_long cmd, caddr_t a break; } error = devfs_set_cdevpriv(mp, ffs_susp_dtor); - KASSERT(error == 0, ("devfs_set_cdevpriv failed")); + if (error != 0) + ffs_susp_unsuspend(mp); break; case UFSRESUME: error = devfs_get_cdevpriv((void **)&mp);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201808011906.w71J6tBe032796>