2012-10-30 20:50:57

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 00/10] NFSD: Improve fault injection

From: Bryan Schumaker <[email protected]>

While working on p2p-nfs, I discovered that I sometimes need to clear
state for a specific client to test all possible error recovery conditions.
The current fault injection code deletes state as it find it, so it can
be difficult to guess which state will be forgotten. In addition, I
currently print out the amount of state forgotten but I don't give details
like "Forgot 3 locks from client w.x.y.z". These patches set out to
improve that.

The first 7 patches clean up the current code and prepare it for specific
client state removal. Patch 8 adds prints information to the server's logs
when a fault injection file is read (such as "Client w.x.y.z has 3 open
files"). Patch 9 adds in a custom file operations structure so users can
write strings to fault injection files in addition to u64s. Finally, patch
10 allows users to remove state by writing a client's IP address to one of
the files.

Changes in v2:
- Patches apply cleanly to the most recent nfsd-next branch
- Improve printk() output

Questions, comments and suggestions are appreciated!

- Bryan

Bryan Schumaker (10):
NFSD: Fold fault_inject.h into state.h
NFSD: Lock state before calling fault injection function
NFSD: Clean up forgetting clients
NFSD: Clean up forgetting locks
NFSD: Clean up forgetting openowners
NFSD: Clean up forgetting and recalling delegations
NFSD: Fault injection operations take a per-client forget function
NFSD: Reading a fault injection file prints a state count
NFSD: Add a custom file operations structure for fault injection
NFSD: Forget state for a specific client

fs/nfsd/fault_inject.c | 112 ++++++++++++++++++++++++++----
fs/nfsd/fault_inject.h | 28 --------
fs/nfsd/nfs4state.c | 181 +++++++++++++++++++++++++++++++------------------
fs/nfsd/nfsctl.c | 2 +-
fs/nfsd/state.h | 23 +++++++
5 files changed, 235 insertions(+), 111 deletions(-)
delete mode 100644 fs/nfsd/fault_inject.h

--
1.8.0



2012-10-30 20:51:02

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 06/10] NFSD: Clean up forgetting and recalling delegations

From: Bryan Schumaker <[email protected]>

Once I have a client, I can easily use its delegation list rather than
searching the file hash table for delegations to remove.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/nfs4state.c | 94 ++++++++++++++++++++++++++++-------------------------
1 file changed, 50 insertions(+), 44 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 552925e..49ab5c4 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4595,6 +4595,52 @@ static u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
return count;
}

+static u64 nfsd_find_n_delegations(struct nfs4_client *clp, u64 max,
+ struct list_head *victims)
+{
+ struct nfs4_delegation *dp, *next;
+ u64 count = 0;
+
+ list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
+ list_move(&dp->dl_recall_lru, victims);
+ if (++count == max)
+ break;
+ }
+ return count;
+}
+
+static u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
+{
+ struct nfs4_delegation *dp, *next;
+ LIST_HEAD(victims);
+ u64 count;
+
+ spin_lock(&recall_lock);
+ count = nfsd_find_n_delegations(clp, max, &victims);
+ spin_unlock(&recall_lock);
+
+ list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
+ unhash_delegation(dp);
+
+ return count;
+}
+
+static u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
+{
+ struct nfs4_delegation *dp, *next;
+ LIST_HEAD(victims);
+ u64 count;
+
+ spin_lock(&recall_lock);
+
+ count = nfsd_find_n_delegations(clp, max, &victims);
+ list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
+ nfsd_break_one_deleg(dp);
+
+ spin_unlock(&recall_lock);
+ return count;
+}
+
static u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
{
struct nfs4_client *clp, *next;
@@ -4627,56 +4673,16 @@ void nfsd_forget_openowners(u64 num)
printk(KERN_INFO "NFSD: Forgot %llu open owners", count);
}

-static int nfsd_process_n_delegations(u64 num, struct list_head *list)
-{
- int i, count = 0;
- struct nfs4_file *fp, *fnext;
- struct nfs4_delegation *dp, *dnext;
-
- for (i = 0; i < FILE_HASH_SIZE; i++) {
- list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
- list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
- list_move(&dp->dl_recall_lru, list);
- if (++count == num)
- return count;
- }
- }
- }
-
- return count;
-}
-
void nfsd_forget_delegations(u64 num)
{
- unsigned int count;
- LIST_HEAD(victims);
- struct nfs4_delegation *dp, *dnext;
-
- spin_lock(&recall_lock);
- count = nfsd_process_n_delegations(num, &victims);
- spin_unlock(&recall_lock);
-
- list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
- unhash_delegation(dp);
-
- printk(KERN_INFO "NFSD: Forgot %d delegations", count);
+ u64 count = nfsd_for_n_state(num, nfsd_forget_client_delegations);
+ printk(KERN_INFO "NFSD: Forgot %llu delegations", count);
}

void nfsd_recall_delegations(u64 num)
{
- unsigned int count;
- LIST_HEAD(victims);
- struct nfs4_delegation *dp, *dnext;
-
- spin_lock(&recall_lock);
- count = nfsd_process_n_delegations(num, &victims);
- list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
- list_del(&dp->dl_recall_lru);
- nfsd_break_one_deleg(dp);
- }
- spin_unlock(&recall_lock);
-
- printk(KERN_INFO "NFSD: Recalled %d delegations", count);
+ u64 count = nfsd_for_n_state(num, nfsd_recall_client_delegations);
+ printk(KERN_INFO "NFSD: Recalled %llu delegations", count);
}

#endif /* CONFIG_NFSD_FAULT_INJECTION */
--
1.8.0


2012-10-30 20:50:59

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 03/10] NFSD: Clean up forgetting clients

From: Bryan Schumaker <[email protected]>

I added in a generic for-each loop that takes a pass over the client_lru
list and calls some function. The next few patches will update other
operations to use this function as well. A value of 0 still means "forget
everything that is found".

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/nfs4state.c | 23 ++++++++++++++++++-----
1 file changed, 18 insertions(+), 5 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index cf79c3b..3c5719a 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4556,18 +4556,31 @@ nfs4_check_open_reclaim(clientid_t *clid, bool sessions)

#ifdef CONFIG_NFSD_FAULT_INJECTION

-void nfsd_forget_clients(u64 num)
+static u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
+{
+ nfsd4_client_record_remove(clp);
+ expire_client(clp);
+ return 1;
+}
+
+static u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
{
struct nfs4_client *clp, *next;
- int count = 0;
+ u64 count = 0;

list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
- expire_client(clp);
- if (++count == num)
+ count += func(clp, max - count);
+ if ((max != 0) && (count >= max))
break;
}

- printk(KERN_INFO "NFSD: Forgot %d clients", count);
+ return count;
+}
+
+void nfsd_forget_clients(u64 num)
+{
+ u64 count = nfsd_for_n_state(num, nfsd_forget_client);
+ printk(KERN_INFO "NFSD: Forgot %llu clients", count);
}

static void release_lockowner_sop(struct nfs4_stateowner *sop)
--
1.8.0


2012-10-30 20:51:01

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 05/10] NFSD: Clean up forgetting openowners

From: Bryan Schumaker <[email protected]>

Using "forget_n_state()" forces me to implement the code needed to
forget a specific client's openowners.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/nfs4state.c | 41 ++++++++++++++++-------------------------
1 file changed, 16 insertions(+), 25 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 5c62323..552925e 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4581,6 +4581,20 @@ static u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
return count;
}

+static u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
+{
+ struct nfs4_openowner *oop, *next;
+ u64 count = 0;
+
+ list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
+ release_openowner(oop);
+ if (++count == max)
+ break;
+ }
+
+ return count;
+}
+
static u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
{
struct nfs4_client *clp, *next;
@@ -4601,29 +4615,6 @@ void nfsd_forget_clients(u64 num)
printk(KERN_INFO "NFSD: Forgot %llu clients", count);
}

-static void release_openowner_sop(struct nfs4_stateowner *sop)
-{
- release_openowner(openowner(sop));
-}
-
-static int nfsd_release_n_owners(u64 num, bool is_open_owner,
- void (*release_sop)(struct nfs4_stateowner *))
-{
- int i, count = 0;
- struct nfs4_stateowner *sop, *next;
-
- for (i = 0; i < OWNER_HASH_SIZE; i++) {
- list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
- if (sop->so_is_open_owner != is_open_owner)
- continue;
- release_sop(sop);
- if (++count == num)
- return count;
- }
- }
- return count;
-}
-
void nfsd_forget_locks(u64 num)
{
u64 count = nfsd_for_n_state(num, nfsd_forget_client_locks);
@@ -4632,8 +4623,8 @@ void nfsd_forget_locks(u64 num)

void nfsd_forget_openowners(u64 num)
{
- int count = nfsd_release_n_owners(num, true, release_openowner_sop);
- printk(KERN_INFO "NFSD: Forgot %d open owners", count);
+ u64 count = nfsd_for_n_state(num, nfsd_forget_client_openowners);
+ printk(KERN_INFO "NFSD: Forgot %llu open owners", count);
}

static int nfsd_process_n_delegations(u64 num, struct list_head *list)
--
1.8.0


2012-10-30 20:51:04

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 09/10] NFSD: Add a custom file operations structure for fault injection

From: Bryan Schumaker <[email protected]>

Controlling the read and write functions allows me to add in "forget
client w.x.y.z", since we won't be limited to reading and writing only
u64 values.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/fault_inject.c | 56 +++++++++++++++++++++++++++++++++++++++++++-------
1 file changed, 49 insertions(+), 7 deletions(-)

diff --git a/fs/nfsd/fault_inject.c b/fs/nfsd/fault_inject.c
index 545f8e4..19f9094 100644
--- a/fs/nfsd/fault_inject.c
+++ b/fs/nfsd/fault_inject.c
@@ -8,6 +8,7 @@
#include <linux/fs.h>
#include <linux/debugfs.h>
#include <linux/module.h>
+#include <asm/uaccess.h>

#include "state.h"

@@ -48,10 +49,9 @@ static struct nfsd_fault_inject_op inject_ops[] = {
static long int NUM_INJECT_OPS = sizeof(inject_ops) / sizeof(struct nfsd_fault_inject_op);
static struct dentry *debug_dir;

-static int nfsd_inject_set(void *op_ptr, u64 val)
+static void nfsd_inject_set(struct nfsd_fault_inject_op *op, u64 val)
{
u64 count = 0;
- struct nfsd_fault_inject_op *op = op_ptr;

if (val == 0)
printk(KERN_INFO "NFSD Fault Injection: %s (all)", op->file);
@@ -62,19 +62,61 @@ static int nfsd_inject_set(void *op_ptr, u64 val)
count = nfsd_for_n_state(val, op->forget);
nfs4_unlock_state();
printk(KERN_INFO "NFSD: %s: found %llu", op->file, count);
- return 0;
}

-static int nfsd_inject_get(void *op_ptr, u64 *val)
+static void nfsd_inject_get(struct nfsd_fault_inject_op *op, u64 *val)
{
- struct nfsd_fault_inject_op *op = op_ptr;
nfs4_lock_state();
*val = nfsd_for_n_state(0, op->print);
nfs4_unlock_state();
- return 0;
}

-DEFINE_SIMPLE_ATTRIBUTE(fops_nfsd, nfsd_inject_get, nfsd_inject_set, "%llu\n");
+static ssize_t fault_inject_read(struct file *file, char __user *buf,
+ size_t len, loff_t *ppos)
+{
+ static u64 val;
+ char read_buf[25];
+ size_t size, ret;
+ loff_t pos = *ppos;
+
+ if (!pos)
+ nfsd_inject_get(file->f_dentry->d_inode->i_private, &val);
+ size = scnprintf(read_buf, sizeof(read_buf), "%llu\n", val);
+
+ if (pos < 0)
+ return -EINVAL;
+ if (pos >= size || !len)
+ return 0;
+ if (len > size - pos)
+ len = size - pos;
+ ret = copy_to_user(buf, read_buf + pos, len);
+ if (ret == len)
+ return -EFAULT;
+ len -= ret;
+ *ppos = pos + len;
+ return len;
+}
+
+static ssize_t fault_inject_write(struct file *file, const char __user *buf,
+ size_t len, loff_t *ppos)
+{
+ char write_buf[24];
+ size_t size = min(sizeof(write_buf), len) - 1;
+ u64 val;
+
+ if (copy_from_user(write_buf, buf, size))
+ return -EFAULT;
+
+ val = simple_strtoll(write_buf, NULL, 0);
+ nfsd_inject_set(file->f_dentry->d_inode->i_private, val);
+ return len; /* on success, claim we got the whole input */
+}
+
+static const struct file_operations fops_nfsd = {
+ .owner = THIS_MODULE,
+ .read = fault_inject_read,
+ .write = fault_inject_write,
+};

void nfsd_fault_inject_cleanup(void)
{
--
1.8.0


2012-10-30 20:51:00

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 04/10] NFSD: Clean up forgetting locks

From: Bryan Schumaker <[email protected]>

I use the new "forget_n_state()" function to iterate through each client
first when searching for locks. This may slow down forgetting locks a
little bit, but it implements most of the code needed to forget a
specified client's locks.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/nfs4state.c | 27 ++++++++++++++++++++-------
1 file changed, 20 insertions(+), 7 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 3c5719a..5c62323 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4563,6 +4563,24 @@ static u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
return 1;
}

+static u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
+{
+ struct nfs4_stateowner *sop, *next;
+ u64 count = 0;
+ int i;
+
+ for (i = 0; i < OWNER_HASH_SIZE; i++) {
+ list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
+ if (!sop->so_is_open_owner && (sop->so_client == clp)) {
+ release_lockowner(lockowner(sop));
+ count++;
+ }
+ }
+ }
+
+ return count;
+}
+
static u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
{
struct nfs4_client *clp, *next;
@@ -4583,11 +4601,6 @@ void nfsd_forget_clients(u64 num)
printk(KERN_INFO "NFSD: Forgot %llu clients", count);
}

-static void release_lockowner_sop(struct nfs4_stateowner *sop)
-{
- release_lockowner(lockowner(sop));
-}
-
static void release_openowner_sop(struct nfs4_stateowner *sop)
{
release_openowner(openowner(sop));
@@ -4613,8 +4626,8 @@ static int nfsd_release_n_owners(u64 num, bool is_open_owner,

void nfsd_forget_locks(u64 num)
{
- int count = nfsd_release_n_owners(num, false, release_lockowner_sop);
- printk(KERN_INFO "NFSD: Forgot %d locks", count);
+ u64 count = nfsd_for_n_state(num, nfsd_forget_client_locks);
+ printk(KERN_INFO "NFSD: Forgot %llu locks", count);
}

void nfsd_forget_openowners(u64 num)
--
1.8.0


2012-10-30 20:51:05

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 10/10] NFSD: Forget state for a specific client

From: Bryan Schumaker <[email protected]>

Write the client's ip address to any state file and all appropriate
state for that client will be forgotten.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/fault_inject.c | 36 ++++++++++++++++++++++++++++++++----
fs/nfsd/nfs4state.c | 11 +++++++++++
fs/nfsd/state.h | 1 +
3 files changed, 44 insertions(+), 4 deletions(-)

diff --git a/fs/nfsd/fault_inject.c b/fs/nfsd/fault_inject.c
index 19f9094..d9c2ef8 100644
--- a/fs/nfsd/fault_inject.c
+++ b/fs/nfsd/fault_inject.c
@@ -8,6 +8,8 @@
#include <linux/fs.h>
#include <linux/debugfs.h>
#include <linux/module.h>
+#include <linux/nsproxy.h>
+#include <linux/sunrpc/clnt.h>
#include <asm/uaccess.h>

#include "state.h"
@@ -64,6 +66,24 @@ static void nfsd_inject_set(struct nfsd_fault_inject_op *op, u64 val)
printk(KERN_INFO "NFSD: %s: found %llu", op->file, count);
}

+static void nfsd_inject_set_client(struct nfsd_fault_inject_op *op,
+ struct sockaddr_storage *addr,
+ size_t addr_size)
+{
+ char buf[INET6_ADDRSTRLEN];
+ struct nfs4_client *clp;
+ u64 count;
+
+ nfs4_lock_state();
+ clp = nfsd_find_client(addr, addr_size);
+ if (clp) {
+ count = op->forget(clp, 0);
+ rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, 129);
+ printk(KERN_INFO "NFSD [%s]: Client %s had %llu state object(s)\n", op->file, buf, count);
+ }
+ nfs4_unlock_state();
+}
+
static void nfsd_inject_get(struct nfsd_fault_inject_op *op, u64 *val)
{
nfs4_lock_state();
@@ -100,15 +120,23 @@ static ssize_t fault_inject_read(struct file *file, char __user *buf,
static ssize_t fault_inject_write(struct file *file, const char __user *buf,
size_t len, loff_t *ppos)
{
- char write_buf[24];
+ char write_buf[INET6_ADDRSTRLEN];
size_t size = min(sizeof(write_buf), len) - 1;
+ struct net *net = current->nsproxy->net_ns;
+ struct sockaddr_storage sa;
u64 val;

if (copy_from_user(write_buf, buf, size))
return -EFAULT;
-
- val = simple_strtoll(write_buf, NULL, 0);
- nfsd_inject_set(file->f_dentry->d_inode->i_private, val);
+ write_buf[size] = '\0';
+
+ size = rpc_pton(net, write_buf, size, (struct sockaddr *)&sa, sizeof(sa));
+ if (size > 0)
+ nfsd_inject_set_client(file->f_dentry->d_inode->i_private, &sa, size);
+ else {
+ val = simple_strtoll(write_buf, NULL, 0);
+ nfsd_inject_set(file->f_dentry->d_inode->i_private, val);
+ }
return len; /* on success, claim we got the whole input */
}

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 0a20ac2..ad02045 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4713,6 +4713,17 @@ u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
return count;
}

+struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
+{
+ struct nfs4_client *clp;
+
+ list_for_each_entry(clp, &client_lru, cl_lru) {
+ if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
+ return clp;
+ }
+ return NULL;
+}
+
#endif /* CONFIG_NFSD_FAULT_INJECTION */

/* initialization to perform at module load time: */
diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
index 646275c..b836e64 100644
--- a/fs/nfsd/state.h
+++ b/fs/nfsd/state.h
@@ -487,6 +487,7 @@ extern void nfsd4_record_grace_done(struct net *net, time_t boot_time);
int nfsd_fault_inject_init(void);
void nfsd_fault_inject_cleanup(void);
u64 nfsd_for_n_state(u64, u64 (*)(struct nfs4_client *, u64));
+struct nfs4_client *nfsd_find_client(struct sockaddr_storage *, size_t);

u64 nfsd_forget_client(struct nfs4_client *, u64);
u64 nfsd_forget_client_locks(struct nfs4_client*, u64);
--
1.8.0


2012-10-30 20:51:04

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 08/10] NFSD: Reading a fault injection file prints a state count

From: Bryan Schumaker <[email protected]>

I also log basic information that I can figure out about the type of
state (such as number of locks for each client IP address). This can be
useful for checking that state was actually dropped and later for
checking if the client was able to recover.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/fault_inject.c | 13 +++++++++--
fs/nfsd/nfs4state.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++
fs/nfsd/state.h | 5 +++++
3 files changed, 74 insertions(+), 2 deletions(-)

diff --git a/fs/nfsd/fault_inject.c b/fs/nfsd/fault_inject.c
index bf6161a..545f8e4 100644
--- a/fs/nfsd/fault_inject.c
+++ b/fs/nfsd/fault_inject.c
@@ -14,28 +14,34 @@
struct nfsd_fault_inject_op {
char *file;
u64 (*forget)(struct nfs4_client *, u64);
+ u64 (*print)(struct nfs4_client *, u64);
};

static struct nfsd_fault_inject_op inject_ops[] = {
{
.file = "forget_clients",
.forget = nfsd_forget_client,
+ .print = nfsd_print_client,
},
{
.file = "forget_locks",
.forget = nfsd_forget_client_locks,
+ .print = nfsd_print_client_locks,
},
{
.file = "forget_openowners",
.forget = nfsd_forget_client_openowners,
+ .print = nfsd_print_client_openowners,
},
{
.file = "forget_delegations",
.forget = nfsd_forget_client_delegations,
+ .print = nfsd_print_client_delegations,
},
{
.file = "recall_delegations",
.forget = nfsd_recall_client_delegations,
+ .print = nfsd_print_client_delegations,
},
};

@@ -59,9 +65,12 @@ static int nfsd_inject_set(void *op_ptr, u64 val)
return 0;
}

-static int nfsd_inject_get(void *data, u64 *val)
+static int nfsd_inject_get(void *op_ptr, u64 *val)
{
- *val = 0;
+ struct nfsd_fault_inject_op *op = op_ptr;
+ nfs4_lock_state();
+ *val = nfsd_for_n_state(0, op->print);
+ nfs4_unlock_state();
return 0;
}

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 3c64dbc..0a20ac2 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4556,6 +4556,22 @@ nfs4_check_open_reclaim(clientid_t *clid, bool sessions)

#ifdef CONFIG_NFSD_FAULT_INJECTION

+u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
+{
+ char buf[INET6_ADDRSTRLEN];
+ rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, 129);
+ printk(KERN_INFO "NFS Client: %s\n", buf);
+ return 1;
+}
+
+static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
+ const char *type)
+{
+ char buf[INET6_ADDRSTRLEN];
+ rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, 129);
+ printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
+}
+
u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
{
nfsd4_client_record_remove(clp);
@@ -4581,6 +4597,23 @@ u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
return count;
}

+u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
+{
+ struct nfs4_stateowner *sop;
+ u64 count = 0;
+ int i;
+
+ for (i = 0; i < OWNER_HASH_SIZE; i++) {
+ list_for_each_entry(sop, &ownerstr_hashtbl[i], so_strhash) {
+ if (!sop->so_is_open_owner && (sop->so_client == clp))
+ count++;
+ }
+ }
+
+ nfsd_print_count(clp, count, "locked files");
+ return count;
+}
+
u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
{
struct nfs4_openowner *oop, *next;
@@ -4595,6 +4628,17 @@ u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
return count;
}

+u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
+{
+ struct nfs4_openowner *oop;
+ u64 count = 0;
+
+ list_for_each_entry(oop, &clp->cl_openowners, oo_perclient)
+ count++;
+ nfsd_print_count(clp, count, "open files");
+ return count;
+}
+
static u64 nfsd_find_n_delegations(struct nfs4_client *clp, u64 max,
struct list_head *victims)
{
@@ -4641,6 +4685,20 @@ u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
return count;
}

+u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
+{
+ struct nfs4_delegation *dp;
+ u64 count = 0;
+
+ spin_lock(&recall_lock);
+ list_for_each_entry(dp, &clp->cl_delegations, dl_perclnt)
+ count++;
+ spin_unlock(&recall_lock);
+
+ nfsd_print_count(clp, count, "delegations");
+ return count;
+}
+
u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
{
struct nfs4_client *clp, *next;
diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
index 9e133e5..646275c 100644
--- a/fs/nfsd/state.h
+++ b/fs/nfsd/state.h
@@ -493,6 +493,11 @@ u64 nfsd_forget_client_locks(struct nfs4_client*, u64);
u64 nfsd_forget_client_openowners(struct nfs4_client *, u64);
u64 nfsd_forget_client_delegations(struct nfs4_client *, u64);
u64 nfsd_recall_client_delegations(struct nfs4_client *, u64);
+
+u64 nfsd_print_client(struct nfs4_client *, u64);
+u64 nfsd_print_client_locks(struct nfs4_client *, u64);
+u64 nfsd_print_client_openowners(struct nfs4_client *, u64);
+u64 nfsd_print_client_delegations(struct nfs4_client *, u64);
#else /* CONFIG_NFSD_FAULT_INJECTION */
static inline int nfsd_fault_inject_init(void) { return 0; }
static inline void nfsd_fault_inject_cleanup(void) {}
--
1.8.0


2012-10-30 20:50:58

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 01/10] NFSD: Fold fault_inject.h into state.h

From: Bryan Schumaker <[email protected]>

There were only a small number of functions in this file and since they
all affect stored state I think it makes sense to put them in state.h
instead. I also dropped most static inline declarations since there are
no callers when fault injection is not enabled.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/fault_inject.c | 1 -
fs/nfsd/fault_inject.h | 28 ----------------------------
fs/nfsd/nfs4state.c | 1 -
fs/nfsd/nfsctl.c | 2 +-
fs/nfsd/state.h | 15 +++++++++++++++
5 files changed, 16 insertions(+), 31 deletions(-)
delete mode 100644 fs/nfsd/fault_inject.h

diff --git a/fs/nfsd/fault_inject.c b/fs/nfsd/fault_inject.c
index e6c3815..0278112 100644
--- a/fs/nfsd/fault_inject.c
+++ b/fs/nfsd/fault_inject.c
@@ -10,7 +10,6 @@
#include <linux/module.h>

#include "state.h"
-#include "fault_inject.h"

struct nfsd_fault_inject_op {
char *file;
diff --git a/fs/nfsd/fault_inject.h b/fs/nfsd/fault_inject.h
deleted file mode 100644
index 90bd057..0000000
--- a/fs/nfsd/fault_inject.h
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2011 Bryan Schumaker <[email protected]>
- *
- * Function definitions for fault injection
- */
-
-#ifndef LINUX_NFSD_FAULT_INJECT_H
-#define LINUX_NFSD_FAULT_INJECT_H
-
-#ifdef CONFIG_NFSD_FAULT_INJECTION
-int nfsd_fault_inject_init(void);
-void nfsd_fault_inject_cleanup(void);
-void nfsd_forget_clients(u64);
-void nfsd_forget_locks(u64);
-void nfsd_forget_openowners(u64);
-void nfsd_forget_delegations(u64);
-void nfsd_recall_delegations(u64);
-#else /* CONFIG_NFSD_FAULT_INJECTION */
-static inline int nfsd_fault_inject_init(void) { return 0; }
-static inline void nfsd_fault_inject_cleanup(void) {}
-static inline void nfsd_forget_clients(u64 num) {}
-static inline void nfsd_forget_locks(u64 num) {}
-static inline void nfsd_forget_openowners(u64 num) {}
-static inline void nfsd_forget_delegations(u64 num) {}
-static inline void nfsd_recall_delegations(u64 num) {}
-#endif /* CONFIG_NFSD_FAULT_INJECTION */
-
-#endif /* LINUX_NFSD_FAULT_INJECT_H */
diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 8fd3ce0..14c3b70 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -44,7 +44,6 @@
#include "xdr4.h"
#include "vfs.h"
#include "current_stateid.h"
-#include "fault_inject.h"

#include "netns.h"

diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index dab350d..7e0d6db 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -19,7 +19,7 @@
#include "idmap.h"
#include "nfsd.h"
#include "cache.h"
-#include "fault_inject.h"
+#include "state.h"
#include "netns.h"

/*
diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
index e036894..d2f52a1 100644
--- a/fs/nfsd/state.h
+++ b/fs/nfsd/state.h
@@ -481,4 +481,19 @@ extern void nfsd4_client_record_create(struct nfs4_client *clp);
extern void nfsd4_client_record_remove(struct nfs4_client *clp);
extern int nfsd4_client_record_check(struct nfs4_client *clp);
extern void nfsd4_record_grace_done(struct net *net, time_t boot_time);
+
+/* nfs fault injection functions */
+#ifdef CONFIG_NFSD_FAULT_INJECTION
+int nfsd_fault_inject_init(void);
+void nfsd_fault_inject_cleanup(void);
+void nfsd_forget_clients(u64);
+void nfsd_forget_locks(u64);
+void nfsd_forget_openowners(u64);
+void nfsd_forget_delegations(u64);
+void nfsd_recall_delegations(u64);
+#else /* CONFIG_NFSD_FAULT_INJECTION */
+static inline int nfsd_fault_inject_init(void) { return 0; }
+static inline void nfsd_fault_inject_cleanup(void) {}
+#endif /* CONFIG_NFSD_FAULT_INJECTION */
+
#endif /* NFSD4_STATE_H */
--
1.8.0


2012-10-30 20:50:58

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 02/10] NFSD: Lock state before calling fault injection function

From: Bryan Schumaker <[email protected]>

Each function touches state in some way, so getting the lock earlier
can help simplify code.

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/fault_inject.c | 2 ++
fs/nfsd/nfs4state.c | 18 ++----------------
2 files changed, 4 insertions(+), 16 deletions(-)

diff --git a/fs/nfsd/fault_inject.c b/fs/nfsd/fault_inject.c
index 0278112..4b385a1 100644
--- a/fs/nfsd/fault_inject.c
+++ b/fs/nfsd/fault_inject.c
@@ -51,7 +51,9 @@ static int nfsd_inject_set(void *op_ptr, u64 val)
else
printk(KERN_INFO "NFSD Fault Injection: %s (n = %llu)", op->file, val);

+ nfs4_lock_state();
op->func(val);
+ nfs4_unlock_state();
return 0;
}

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 14c3b70..cf79c3b 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4561,13 +4561,11 @@ void nfsd_forget_clients(u64 num)
struct nfs4_client *clp, *next;
int count = 0;

- nfs4_lock_state();
list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
expire_client(clp);
if (++count == num)
break;
}
- nfs4_unlock_state();

printk(KERN_INFO "NFSD: Forgot %d clients", count);
}
@@ -4602,23 +4600,13 @@ static int nfsd_release_n_owners(u64 num, bool is_open_owner,

void nfsd_forget_locks(u64 num)
{
- int count;
-
- nfs4_lock_state();
- count = nfsd_release_n_owners(num, false, release_lockowner_sop);
- nfs4_unlock_state();
-
+ int count = nfsd_release_n_owners(num, false, release_lockowner_sop);
printk(KERN_INFO "NFSD: Forgot %d locks", count);
}

void nfsd_forget_openowners(u64 num)
{
- int count;
-
- nfs4_lock_state();
- count = nfsd_release_n_owners(num, true, release_openowner_sop);
- nfs4_unlock_state();
-
+ int count = nfsd_release_n_owners(num, true, release_openowner_sop);
printk(KERN_INFO "NFSD: Forgot %d open owners", count);
}

@@ -4651,10 +4639,8 @@ void nfsd_forget_delegations(u64 num)
count = nfsd_process_n_delegations(num, &victims);
spin_unlock(&recall_lock);

- nfs4_lock_state();
list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
unhash_delegation(dp);
- nfs4_unlock_state();

printk(KERN_INFO "NFSD: Forgot %d delegations", count);
}
--
1.8.0


2012-10-30 20:51:03

by Anna Schumaker

[permalink] [raw]
Subject: [PATCH v2 07/10] NFSD: Fault injection operations take a per-client forget function

From: Bryan Schumaker <[email protected]>

The eventual goal is to forget state based on ip address, so it makes
sense to call this function in a for-each-client loop until the correct
amount of state is forgotten. I also use this patch as an opportunity
to rename the forget function from "func()" to "forget()".

Signed-off-by: Bryan Schumaker <[email protected]>
---
fs/nfsd/fault_inject.c | 16 +++++++++-------
fs/nfsd/nfs4state.c | 42 ++++++------------------------------------
fs/nfsd/state.h | 12 +++++++-----
3 files changed, 22 insertions(+), 48 deletions(-)

diff --git a/fs/nfsd/fault_inject.c b/fs/nfsd/fault_inject.c
index 4b385a1..bf6161a 100644
--- a/fs/nfsd/fault_inject.c
+++ b/fs/nfsd/fault_inject.c
@@ -13,29 +13,29 @@

struct nfsd_fault_inject_op {
char *file;
- void (*func)(u64);
+ u64 (*forget)(struct nfs4_client *, u64);
};

static struct nfsd_fault_inject_op inject_ops[] = {
{
.file = "forget_clients",
- .func = nfsd_forget_clients,
+ .forget = nfsd_forget_client,
},
{
.file = "forget_locks",
- .func = nfsd_forget_locks,
+ .forget = nfsd_forget_client_locks,
},
{
.file = "forget_openowners",
- .func = nfsd_forget_openowners,
+ .forget = nfsd_forget_client_openowners,
},
{
.file = "forget_delegations",
- .func = nfsd_forget_delegations,
+ .forget = nfsd_forget_client_delegations,
},
{
.file = "recall_delegations",
- .func = nfsd_recall_delegations,
+ .forget = nfsd_recall_client_delegations,
},
};

@@ -44,6 +44,7 @@ static struct dentry *debug_dir;

static int nfsd_inject_set(void *op_ptr, u64 val)
{
+ u64 count = 0;
struct nfsd_fault_inject_op *op = op_ptr;

if (val == 0)
@@ -52,8 +53,9 @@ static int nfsd_inject_set(void *op_ptr, u64 val)
printk(KERN_INFO "NFSD Fault Injection: %s (n = %llu)", op->file, val);

nfs4_lock_state();
- op->func(val);
+ count = nfsd_for_n_state(val, op->forget);
nfs4_unlock_state();
+ printk(KERN_INFO "NFSD: %s: found %llu", op->file, count);
return 0;
}

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 49ab5c4..3c64dbc 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -4556,14 +4556,14 @@ nfs4_check_open_reclaim(clientid_t *clid, bool sessions)

#ifdef CONFIG_NFSD_FAULT_INJECTION

-static u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
+u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
{
nfsd4_client_record_remove(clp);
expire_client(clp);
return 1;
}

-static u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
+u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
{
struct nfs4_stateowner *sop, *next;
u64 count = 0;
@@ -4581,7 +4581,7 @@ static u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
return count;
}

-static u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
+u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
{
struct nfs4_openowner *oop, *next;
u64 count = 0;
@@ -4609,7 +4609,7 @@ static u64 nfsd_find_n_delegations(struct nfs4_client *clp, u64 max,
return count;
}

-static u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
+u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
{
struct nfs4_delegation *dp, *next;
LIST_HEAD(victims);
@@ -4625,7 +4625,7 @@ static u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
return count;
}

-static u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
+u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
{
struct nfs4_delegation *dp, *next;
LIST_HEAD(victims);
@@ -4641,7 +4641,7 @@ static u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
return count;
}

-static u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
+u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
{
struct nfs4_client *clp, *next;
u64 count = 0;
@@ -4655,36 +4655,6 @@ static u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
return count;
}

-void nfsd_forget_clients(u64 num)
-{
- u64 count = nfsd_for_n_state(num, nfsd_forget_client);
- printk(KERN_INFO "NFSD: Forgot %llu clients", count);
-}
-
-void nfsd_forget_locks(u64 num)
-{
- u64 count = nfsd_for_n_state(num, nfsd_forget_client_locks);
- printk(KERN_INFO "NFSD: Forgot %llu locks", count);
-}
-
-void nfsd_forget_openowners(u64 num)
-{
- u64 count = nfsd_for_n_state(num, nfsd_forget_client_openowners);
- printk(KERN_INFO "NFSD: Forgot %llu open owners", count);
-}
-
-void nfsd_forget_delegations(u64 num)
-{
- u64 count = nfsd_for_n_state(num, nfsd_forget_client_delegations);
- printk(KERN_INFO "NFSD: Forgot %llu delegations", count);
-}
-
-void nfsd_recall_delegations(u64 num)
-{
- u64 count = nfsd_for_n_state(num, nfsd_recall_client_delegations);
- printk(KERN_INFO "NFSD: Recalled %llu delegations", count);
-}
-
#endif /* CONFIG_NFSD_FAULT_INJECTION */

/* initialization to perform at module load time: */
diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
index d2f52a1..9e133e5 100644
--- a/fs/nfsd/state.h
+++ b/fs/nfsd/state.h
@@ -486,11 +486,13 @@ extern void nfsd4_record_grace_done(struct net *net, time_t boot_time);
#ifdef CONFIG_NFSD_FAULT_INJECTION
int nfsd_fault_inject_init(void);
void nfsd_fault_inject_cleanup(void);
-void nfsd_forget_clients(u64);
-void nfsd_forget_locks(u64);
-void nfsd_forget_openowners(u64);
-void nfsd_forget_delegations(u64);
-void nfsd_recall_delegations(u64);
+u64 nfsd_for_n_state(u64, u64 (*)(struct nfs4_client *, u64));
+
+u64 nfsd_forget_client(struct nfs4_client *, u64);
+u64 nfsd_forget_client_locks(struct nfs4_client*, u64);
+u64 nfsd_forget_client_openowners(struct nfs4_client *, u64);
+u64 nfsd_forget_client_delegations(struct nfs4_client *, u64);
+u64 nfsd_recall_client_delegations(struct nfs4_client *, u64);
#else /* CONFIG_NFSD_FAULT_INJECTION */
static inline int nfsd_fault_inject_init(void) { return 0; }
static inline void nfsd_fault_inject_cleanup(void) {}
--
1.8.0


2012-11-26 15:18:00

by Anna Schumaker

[permalink] [raw]
Subject: Re: [PATCH v2 00/10] NFSD: Improve fault injection

On 11/15/2012 10:00 AM, J. Bruce Fields wrote:
> On Thu, Nov 15, 2012 at 09:55:24AM -0500, Bryan Schumaker wrote:
>> On 11/14/2012 05:48 PM, J. Bruce Fields wrote:
>>> On Tue, Oct 30, 2012 at 04:50:44PM -0400, [email protected] wrote:
>>>> From: Bryan Schumaker <[email protected]>
>>>>
>>>> While working on p2p-nfs, I discovered that I sometimes need to clear
>>>> state for a specific client to test all possible error recovery conditions.
>>>> The current fault injection code deletes state as it find it, so it can
>>>> be difficult to guess which state will be forgotten. In addition, I
>>>> currently print out the amount of state forgotten but I don't give details
>>>> like "Forgot 3 locks from client w.x.y.z". These patches set out to
>>>> improve that.
>>>>
>>>> The first 7 patches clean up the current code and prepare it for specific
>>>> client state removal. Patch 8 adds prints information to the server's logs
>>>> when a fault injection file is read (such as "Client w.x.y.z has 3 open
>>>> files"). Patch 9 adds in a custom file operations structure so users can
>>>> write strings to fault injection files in addition to u64s. Finally, patch
>>>> 10 allows users to remove state by writing a client's IP address to one of
>>>> the files.
>>>
>>> Apologies for putting you off again, but: I'd like to get Stanislav's
>>> patches sorted out first. I think that shouldn't take too long--bug me
>>> in another week or so if I haven't come back to these by then. These
>>> may need a little rebasing at that point.
>>
>> Ok, I'll add "bug Bruce" to my calendar... I took a quick glance through his patches and I don't think rebasing will be too hard.
>
> Good, thanks!

So what's the plan for these patches?

- Bryan
>
> --b.
>


2012-11-15 15:00:42

by J. Bruce Fields

[permalink] [raw]
Subject: Re: [PATCH v2 00/10] NFSD: Improve fault injection

On Thu, Nov 15, 2012 at 09:55:24AM -0500, Bryan Schumaker wrote:
> On 11/14/2012 05:48 PM, J. Bruce Fields wrote:
> > On Tue, Oct 30, 2012 at 04:50:44PM -0400, [email protected] wrote:
> >> From: Bryan Schumaker <[email protected]>
> >>
> >> While working on p2p-nfs, I discovered that I sometimes need to clear
> >> state for a specific client to test all possible error recovery conditions.
> >> The current fault injection code deletes state as it find it, so it can
> >> be difficult to guess which state will be forgotten. In addition, I
> >> currently print out the amount of state forgotten but I don't give details
> >> like "Forgot 3 locks from client w.x.y.z". These patches set out to
> >> improve that.
> >>
> >> The first 7 patches clean up the current code and prepare it for specific
> >> client state removal. Patch 8 adds prints information to the server's logs
> >> when a fault injection file is read (such as "Client w.x.y.z has 3 open
> >> files"). Patch 9 adds in a custom file operations structure so users can
> >> write strings to fault injection files in addition to u64s. Finally, patch
> >> 10 allows users to remove state by writing a client's IP address to one of
> >> the files.
> >
> > Apologies for putting you off again, but: I'd like to get Stanislav's
> > patches sorted out first. I think that shouldn't take too long--bug me
> > in another week or so if I haven't come back to these by then. These
> > may need a little rebasing at that point.
>
> Ok, I'll add "bug Bruce" to my calendar... I took a quick glance through his patches and I don't think rebasing will be too hard.

Good, thanks!

--b.

2012-11-26 16:26:13

by Anna Schumaker

[permalink] [raw]
Subject: Re: [PATCH v2 00/10] NFSD: Improve fault injection

On 11/26/2012 11:10 AM, J. Bruce Fields wrote:
> On Mon, Nov 26, 2012 at 10:17:58AM -0500, Bryan Schumaker wrote:
>> On 11/15/2012 10:00 AM, J. Bruce Fields wrote:
>>> On Thu, Nov 15, 2012 at 09:55:24AM -0500, Bryan Schumaker wrote:
>>>> On 11/14/2012 05:48 PM, J. Bruce Fields wrote:
>>>>> On Tue, Oct 30, 2012 at 04:50:44PM -0400, [email protected] wrote:
>>>>>> From: Bryan Schumaker <[email protected]>
>>>>>>
>>>>>> While working on p2p-nfs, I discovered that I sometimes need to clear
>>>>>> state for a specific client to test all possible error recovery conditions.
>>>>>> The current fault injection code deletes state as it find it, so it can
>>>>>> be difficult to guess which state will be forgotten. In addition, I
>>>>>> currently print out the amount of state forgotten but I don't give details
>>>>>> like "Forgot 3 locks from client w.x.y.z". These patches set out to
>>>>>> improve that.
>>>>>>
>>>>>> The first 7 patches clean up the current code and prepare it for specific
>>>>>> client state removal. Patch 8 adds prints information to the server's logs
>>>>>> when a fault injection file is read (such as "Client w.x.y.z has 3 open
>>>>>> files"). Patch 9 adds in a custom file operations structure so users can
>>>>>> write strings to fault injection files in addition to u64s. Finally, patch
>>>>>> 10 allows users to remove state by writing a client's IP address to one of
>>>>>> the files.
>>>>>
>>>>> Apologies for putting you off again, but: I'd like to get Stanislav's
>>>>> patches sorted out first. I think that shouldn't take too long--bug me
>>>>> in another week or so if I haven't come back to these by then. These
>>>>> may need a little rebasing at that point.
>>>>
>>>> Ok, I'll add "bug Bruce" to my calendar... I took a quick glance through his patches and I don't think rebasing will be too hard.
>>>
>>> Good, thanks!
>>
>> So what's the plan for these patches?
>
> I'll need to take a quick look at Stanislav's latest and at these
> patches. Some of his stuff is already in my public tree so yours may
> already need a rebase?

Yeah, some of the context has changed for these patches... updating now.

- Bryan

>
> --b.
>


2012-11-26 16:10:48

by J. Bruce Fields

[permalink] [raw]
Subject: Re: [PATCH v2 00/10] NFSD: Improve fault injection

On Mon, Nov 26, 2012 at 10:17:58AM -0500, Bryan Schumaker wrote:
> On 11/15/2012 10:00 AM, J. Bruce Fields wrote:
> > On Thu, Nov 15, 2012 at 09:55:24AM -0500, Bryan Schumaker wrote:
> >> On 11/14/2012 05:48 PM, J. Bruce Fields wrote:
> >>> On Tue, Oct 30, 2012 at 04:50:44PM -0400, [email protected] wrote:
> >>>> From: Bryan Schumaker <[email protected]>
> >>>>
> >>>> While working on p2p-nfs, I discovered that I sometimes need to clear
> >>>> state for a specific client to test all possible error recovery conditions.
> >>>> The current fault injection code deletes state as it find it, so it can
> >>>> be difficult to guess which state will be forgotten. In addition, I
> >>>> currently print out the amount of state forgotten but I don't give details
> >>>> like "Forgot 3 locks from client w.x.y.z". These patches set out to
> >>>> improve that.
> >>>>
> >>>> The first 7 patches clean up the current code and prepare it for specific
> >>>> client state removal. Patch 8 adds prints information to the server's logs
> >>>> when a fault injection file is read (such as "Client w.x.y.z has 3 open
> >>>> files"). Patch 9 adds in a custom file operations structure so users can
> >>>> write strings to fault injection files in addition to u64s. Finally, patch
> >>>> 10 allows users to remove state by writing a client's IP address to one of
> >>>> the files.
> >>>
> >>> Apologies for putting you off again, but: I'd like to get Stanislav's
> >>> patches sorted out first. I think that shouldn't take too long--bug me
> >>> in another week or so if I haven't come back to these by then. These
> >>> may need a little rebasing at that point.
> >>
> >> Ok, I'll add "bug Bruce" to my calendar... I took a quick glance through his patches and I don't think rebasing will be too hard.
> >
> > Good, thanks!
>
> So what's the plan for these patches?

I'll need to take a quick look at Stanislav's latest and at these
patches. Some of his stuff is already in my public tree so yours may
already need a rebase?

--b.

2012-11-15 14:55:26

by Anna Schumaker

[permalink] [raw]
Subject: Re: [PATCH v2 00/10] NFSD: Improve fault injection

On 11/14/2012 05:48 PM, J. Bruce Fields wrote:
> On Tue, Oct 30, 2012 at 04:50:44PM -0400, [email protected] wrote:
>> From: Bryan Schumaker <[email protected]>
>>
>> While working on p2p-nfs, I discovered that I sometimes need to clear
>> state for a specific client to test all possible error recovery conditions.
>> The current fault injection code deletes state as it find it, so it can
>> be difficult to guess which state will be forgotten. In addition, I
>> currently print out the amount of state forgotten but I don't give details
>> like "Forgot 3 locks from client w.x.y.z". These patches set out to
>> improve that.
>>
>> The first 7 patches clean up the current code and prepare it for specific
>> client state removal. Patch 8 adds prints information to the server's logs
>> when a fault injection file is read (such as "Client w.x.y.z has 3 open
>> files"). Patch 9 adds in a custom file operations structure so users can
>> write strings to fault injection files in addition to u64s. Finally, patch
>> 10 allows users to remove state by writing a client's IP address to one of
>> the files.
>
> Apologies for putting you off again, but: I'd like to get Stanislav's
> patches sorted out first. I think that shouldn't take too long--bug me
> in another week or so if I haven't come back to these by then. These
> may need a little rebasing at that point.

Ok, I'll add "bug Bruce" to my calendar... I took a quick glance through his patches and I don't think rebasing will be too hard.

- Bryan

>
> --b.
>
>>
>> Changes in v2:
>> - Patches apply cleanly to the most recent nfsd-next branch
>> - Improve printk() output
>>
>> Questions, comments and suggestions are appreciated!
>>
>> - Bryan
>>
>> Bryan Schumaker (10):
>> NFSD: Fold fault_inject.h into state.h
>> NFSD: Lock state before calling fault injection function
>> NFSD: Clean up forgetting clients
>> NFSD: Clean up forgetting locks
>> NFSD: Clean up forgetting openowners
>> NFSD: Clean up forgetting and recalling delegations
>> NFSD: Fault injection operations take a per-client forget function
>> NFSD: Reading a fault injection file prints a state count
>> NFSD: Add a custom file operations structure for fault injection
>> NFSD: Forget state for a specific client
>>
>> fs/nfsd/fault_inject.c | 112 ++++++++++++++++++++++++++----
>> fs/nfsd/fault_inject.h | 28 --------
>> fs/nfsd/nfs4state.c | 181 +++++++++++++++++++++++++++++++------------------
>> fs/nfsd/nfsctl.c | 2 +-
>> fs/nfsd/state.h | 23 +++++++
>> 5 files changed, 235 insertions(+), 111 deletions(-)
>> delete mode 100644 fs/nfsd/fault_inject.h
>>
>> --
>> 1.8.0
>>


2012-11-14 22:48:48

by J. Bruce Fields

[permalink] [raw]
Subject: Re: [PATCH v2 00/10] NFSD: Improve fault injection

On Tue, Oct 30, 2012 at 04:50:44PM -0400, [email protected] wrote:
> From: Bryan Schumaker <[email protected]>
>
> While working on p2p-nfs, I discovered that I sometimes need to clear
> state for a specific client to test all possible error recovery conditions.
> The current fault injection code deletes state as it find it, so it can
> be difficult to guess which state will be forgotten. In addition, I
> currently print out the amount of state forgotten but I don't give details
> like "Forgot 3 locks from client w.x.y.z". These patches set out to
> improve that.
>
> The first 7 patches clean up the current code and prepare it for specific
> client state removal. Patch 8 adds prints information to the server's logs
> when a fault injection file is read (such as "Client w.x.y.z has 3 open
> files"). Patch 9 adds in a custom file operations structure so users can
> write strings to fault injection files in addition to u64s. Finally, patch
> 10 allows users to remove state by writing a client's IP address to one of
> the files.

Apologies for putting you off again, but: I'd like to get Stanislav's
patches sorted out first. I think that shouldn't take too long--bug me
in another week or so if I haven't come back to these by then. These
may need a little rebasing at that point.

--b.

>
> Changes in v2:
> - Patches apply cleanly to the most recent nfsd-next branch
> - Improve printk() output
>
> Questions, comments and suggestions are appreciated!
>
> - Bryan
>
> Bryan Schumaker (10):
> NFSD: Fold fault_inject.h into state.h
> NFSD: Lock state before calling fault injection function
> NFSD: Clean up forgetting clients
> NFSD: Clean up forgetting locks
> NFSD: Clean up forgetting openowners
> NFSD: Clean up forgetting and recalling delegations
> NFSD: Fault injection operations take a per-client forget function
> NFSD: Reading a fault injection file prints a state count
> NFSD: Add a custom file operations structure for fault injection
> NFSD: Forget state for a specific client
>
> fs/nfsd/fault_inject.c | 112 ++++++++++++++++++++++++++----
> fs/nfsd/fault_inject.h | 28 --------
> fs/nfsd/nfs4state.c | 181 +++++++++++++++++++++++++++++++------------------
> fs/nfsd/nfsctl.c | 2 +-
> fs/nfsd/state.h | 23 +++++++
> 5 files changed, 235 insertions(+), 111 deletions(-)
> delete mode 100644 fs/nfsd/fault_inject.h
>
> --
> 1.8.0
>