Skip to contentSkip to conversation

fs: dlm: dlm lock from user space fixes

PATCH New Alexander Aring 1 participant 3 messages Dec 11, 2020 Distributed lock manager (DLM)

lore
Share & tools
lore

Alexander Aring

author maintainer
#0 lore raw

cover fs: dlm: dlm lock from user space fixes

Hi,

found some issues with the user space fixes and that we don't have proper aligned dlm messages. I found these issues while having my pending patches applied which introduced a "hart" assert for this and dropping messages on such case. With this assert dlm will probably not backwards compatible anymore...

so here my question/suggestion:

We should queue these patches to stable?

- Alex

btw: these patches can be applied, I will send PATCHv3 for my pending patches, I added two other patches.

Alexander Aring (2):
fs: dlm: check for invalid namelen
fs: dlm: transparently align name buffer

 fs/dlm/user.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
Signature · 1 line
2.26.2
author maintainer
#1 lore raw

PATCH 1/2 fs: dlm: check for invalid namelen

This patch adds an additional check inside the dlm locking from user space functionality that the namelen isn't above the maximum allowed dlm resource name length.

---
 fs/dlm/user.c | 3 +++
 1 file changed, 3 insertions(+)
fs/dlm/user.c +3 -0
index e5cefa90b1ce..9f2f743eeb31 100644
@@ -241,6 +241,9 @@ static int device_user_lock(struct dlm_user_proc *proc,
 	uint32_t lkid;
 	int error = -ENOMEM;
 
+	if (params->namelen > DLM_RESNAME_MAXLEN)
+		return -EINVAL;
+
 	ls = dlm_find_lockspace_local(proc->lockspace);
 	if (!ls)
 		return -ENOENT;
Signature · 1 line
2.26.2
author maintainer
#2 lore raw

PATCH 2/2 fs: dlm: transparently align name buffer

In reply to Alexander Aring

This patch adds a zeroed temporary buffer to align the given resource name of a dlm lock user space request transparently to a 4 byte aligned buffer. I hit the assert while running the recent pending patch which contains a 4 byte aligned assert for the dlm header msglength field. It's important that the allocation and msglength field is 4 byte aligned as the dlm_header structure is natural aligned to 4 byte values. The internal dlm allocator can return the next remaining space from a previously allocated length. The buffer will not be aligned to 4 byte anymore if this case occurs. At receiving side we cannot parse the next dlm message because the next dlm header starts at a unaligned address as well.

---
 fs/dlm/user.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
fs/dlm/user.c +5 -2
index 9f2f743eeb31..70a5116df9fe 100644
@@ -236,6 +236,7 @@ void dlm_user_add_ast(struct dlm_lkb *lkb, uint32_t flags, int mode,
 static int device_user_lock(struct dlm_user_proc *proc,
 			    struct dlm_lock_params *params)
 {
+	unsigned char name[DLM_RESNAME_MAXLEN] = {};
 	struct dlm_ls *ls;
 	struct dlm_user_args *ua;
 	uint32_t lkid;
@@ -270,17 +271,19 @@ static int device_user_lock(struct dlm_user_proc *proc,
 				         params->lkid, params->lvb,
 					 (unsigned long) params->timeout);
 	} else if (params->flags & DLM_LKF_ORPHAN) {
+		memcpy(name, params->name, params->namelen);
 		error = dlm_user_adopt_orphan(ls, ua,
 					 params->mode, params->flags,
-					 params->name, params->namelen,
+					 name, ALIGN(params->namelen, 4),
 					 (unsigned long) params->timeout,
 					 &lkid);
 		if (!error)
 			error = lkid;
 	} else {
+		memcpy(name, params->name, params->namelen);
 		error = dlm_user_request(ls, ua,
 					 params->mode, params->flags,
-					 params->name, params->namelen,
+					 name, ALIGN(params->namelen, 4),
 					 (unsigned long) params->timeout);
 		if (!error)
 			error = ua->lksb.sb_lkid;
Signature · 1 line
2.26.2