From 3d2fec7db856c67e1a94a87a846d8ffe6f48b61f Mon Sep 17 00:00:00 2001
From: Dmitry Chagin <dchagin@FreeBSD.org>
Date: Mon, 29 May 2023 11:15:28 +0300
Subject: namei: Add the abilty for the ABI to specify an alternate root path

For now a non-native ABI (i.e., Linux) uses the kern_alternate_path()
facility to dynamically reroot lookups. First, an attempt is made to
lookup the file in /compat/linux/original-path. If that fails, the
lookup is done in /original-path. Thats requires a bit of code in
every ABI syscall implementation where path name translation is needed.
Also our kern_alternate_path() does not properly lookups absolute symlinks
in second attempt, i.e., does not append /compat/linux part to the resolved
link.
The change is intended to avoid this by specifiyng the ABI root directory
for namei(), using one call to pwd_altroot() during exec-time into the ABI.
In that case namei() will dynamically reroot lookups as mentioned above.

PR:			72920
Reviewed by:		kib
Differential revision:	https://reviews.freebsd.org/D38933
MFC after:		2 month
---
 sys/kern/kern_descrip.c | 61 +++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 59 insertions(+), 2 deletions(-)

(limited to 'sys/kern/kern_descrip.c')

diff --git a/sys/kern/kern_descrip.c b/sys/kern/kern_descrip.c
index 0be59e930dd4..908c3352514b 100644
--- a/sys/kern/kern_descrip.c
+++ b/sys/kern/kern_descrip.c
@@ -3839,6 +3839,11 @@ pwd_fill(struct pwd *oldpwd, struct pwd *newpwd)
 		vrefact(oldpwd->pwd_jdir);
 		newpwd->pwd_jdir = oldpwd->pwd_jdir;
 	}
+
+	if (newpwd->pwd_adir == NULL && oldpwd->pwd_adir != NULL) {
+		vrefact(oldpwd->pwd_adir);
+		newpwd->pwd_adir = oldpwd->pwd_adir;
+	}
 }
 
 struct pwd *
@@ -3930,6 +3935,8 @@ pwd_drop(struct pwd *pwd)
 		vrele(pwd->pwd_rdir);
 	if (pwd->pwd_jdir != NULL)
 		vrele(pwd->pwd_jdir);
+	if (pwd->pwd_adir != NULL)
+		vrele(pwd->pwd_adir);
 	uma_zfree_smr(pwd_zone, pwd);
 }
 
@@ -3967,6 +3974,8 @@ pwd_chroot(struct thread *td, struct vnode *vp)
 
 	vrefact(vp);
 	newpwd->pwd_rdir = vp;
+	vrefact(vp);
+	newpwd->pwd_adir = vp;
 	if (oldpwd->pwd_jdir == NULL) {
 		vrefact(vp);
 		newpwd->pwd_jdir = vp;
@@ -3997,6 +4006,40 @@ pwd_chdir(struct thread *td, struct vnode *vp)
 	pwd_drop(oldpwd);
 }
 
+/*
+ * Process is transitioning to/from a non-native ABI.
+ */
+void
+pwd_altroot(struct thread *td, struct vnode *altroot_vp)
+{
+	struct pwddesc *pdp;
+	struct pwd *newpwd, *oldpwd;
+
+	newpwd = pwd_alloc();
+	pdp = td->td_proc->p_pd;
+	PWDDESC_XLOCK(pdp);
+	oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
+	if (altroot_vp != NULL) {
+		/*
+		 * Native process to a non-native ABI.
+		 */
+
+		vrefact(altroot_vp);
+		newpwd->pwd_adir = altroot_vp;
+	} else {
+		/*
+		 * Non-native process to the native ABI.
+		 */
+
+		vrefact(oldpwd->pwd_rdir);
+		newpwd->pwd_adir = oldpwd->pwd_rdir;
+	}
+	pwd_fill(oldpwd, newpwd);
+	pwd_set(pdp, newpwd);
+	PWDDESC_XUNLOCK(pdp);
+	pwd_drop(oldpwd);
+}
+
 /*
  * jail_attach(2) changes both root and working directories.
  */
@@ -4030,6 +4073,8 @@ pwd_chroot_chdir(struct thread *td, struct vnode *vp)
 		vrefact(vp);
 		newpwd->pwd_jdir = vp;
 	}
+	vrefact(vp);
+	newpwd->pwd_adir = vp;
 	pwd_fill(oldpwd, newpwd);
 	pwd_set(pdp, newpwd);
 	PWDDESC_XUNLOCK(pdp);
@@ -4046,7 +4091,8 @@ pwd_ensure_dirs(void)
 	pdp = curproc->p_pd;
 	PWDDESC_XLOCK(pdp);
 	oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
-	if (oldpwd->pwd_cdir != NULL && oldpwd->pwd_rdir != NULL) {
+	if (oldpwd->pwd_cdir != NULL && oldpwd->pwd_rdir != NULL &&
+	    oldpwd->pwd_adir != NULL) {
 		PWDDESC_XUNLOCK(pdp);
 		return;
 	}
@@ -4064,6 +4110,10 @@ pwd_ensure_dirs(void)
 		vrefact(rootvnode);
 		newpwd->pwd_rdir = rootvnode;
 	}
+	if (newpwd->pwd_adir == NULL) {
+		vrefact(rootvnode);
+		newpwd->pwd_adir = rootvnode;
+	}
 	pwd_set(pdp, newpwd);
 	PWDDESC_XUNLOCK(pdp);
 	pwd_drop(oldpwd);
@@ -4084,6 +4134,8 @@ pwd_set_rootvnode(void)
 	newpwd->pwd_cdir = rootvnode;
 	vrefact(rootvnode);
 	newpwd->pwd_rdir = rootvnode;
+	vrefact(rootvnode);
+	newpwd->pwd_adir = rootvnode;
 	pwd_fill(oldpwd, newpwd);
 	pwd_set(pdp, newpwd);
 	PWDDESC_XUNLOCK(pdp);
@@ -4119,7 +4171,8 @@ mountcheckdirs(struct vnode *olddp, struct vnode *newdp)
 		if (oldpwd == NULL ||
 		    (oldpwd->pwd_cdir != olddp &&
 		    oldpwd->pwd_rdir != olddp &&
-		    oldpwd->pwd_jdir != olddp)) {
+		    oldpwd->pwd_jdir != olddp &&
+		    oldpwd->pwd_adir != olddp)) {
 			PWDDESC_XUNLOCK(pdp);
 			pddrop(pdp);
 			continue;
@@ -4136,6 +4189,10 @@ mountcheckdirs(struct vnode *olddp, struct vnode *newdp)
 			vrefact(newdp);
 			newpwd->pwd_jdir = newdp;
 		}
+		if (oldpwd->pwd_adir == olddp) {
+			vrefact(newdp);
+			newpwd->pwd_adir = newdp;
+		}
 		pwd_fill(oldpwd, newpwd);
 		pwd_set(pdp, newpwd);
 		PWDDESC_XUNLOCK(pdp);
-- 
cgit v1.2.3