Fix: update writeback instrumentation for kernel 4.14

[lttng-modules.git] / lttng-statedump-impl.c

diff --git a/lttng-statedump-impl.c b/lttng-statedump-impl.c
index d0e641a97ca70452508ccf618ebf9c594b67cb5b..a2520433216f0f79d5a502cf5dbb4389222a7d0a 100644 (file)
--- a/lttng-statedump-impl.c
+++ b/lttng-statedump-impl.c
@@ -47,15 +47,17 @@
 #include <linux/mutex.h>
 #include <linux/device.h>
 
-#include "lttng-events.h"
-#include "lttng-tracer.h"
-#include "wrapper/irqdesc.h"
-#include "wrapper/spinlock.h"
-#include "wrapper/fdtable.h"
-#include "wrapper/nsproxy.h"
-#include "wrapper/irq.h"
-#include "wrapper/tracepoint.h"
-#include "wrapper/genhd.h"
+#include <lttng-events.h>
+#include <lttng-tracer.h>
+#include <wrapper/irqdesc.h>
+#include <wrapper/spinlock.h>
+#include <wrapper/fdtable.h>
+#include <wrapper/nsproxy.h>
+#include <wrapper/irq.h>
+#include <wrapper/tracepoint.h>
+#include <wrapper/genhd.h>
+#include <wrapper/file.h>
+#include <wrapper/time.h>
 
 #ifdef CONFIG_LTTNG_HAS_LIST_IRQ
 #include <linux/irq.h>
@@ -63,9 +65,10 @@
 
 /* Define the tracepoints, but do not build the probes */
 #define CREATE_TRACE_POINTS
-#define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
+#define TRACE_INCLUDE_PATH instrumentation/events/lttng-module
 #define TRACE_INCLUDE_FILE lttng-statedump
-#include "instrumentation/events/lttng-module/lttng-statedump.h"
+#define LTTNG_INSTRUMENTATION
+#include <instrumentation/events/lttng-module/lttng-statedump.h>
 
 DEFINE_TRACE(lttng_statedump_block_device);
 DEFINE_TRACE(lttng_statedump_end);
@@ -79,7 +82,7 @@ struct lttng_fd_ctx {
        char *page;
        struct lttng_session *session;
        struct task_struct *p;
-       struct fdtable *fdt;
+       struct files_struct *files;
 };
 
 /*
@@ -219,13 +222,24 @@ int lttng_dump_one_fd(const void *p, struct file *file, unsigned int fd)
        const struct lttng_fd_ctx *ctx = p;
        const char *s = d_path(&file->f_path, ctx->page, PAGE_SIZE);
        unsigned int flags = file->f_flags;
+       struct fdtable *fdt;
 
        /*
         * We don't expose kernel internal flags, only userspace-visible
         * flags.
         */
        flags &= ~FMODE_NONOTIFY;
-       if (test_bit(fd, ctx->fdt->close_on_exec))
+       fdt = files_fdtable(ctx->files);
+       /*
+        * We need to check here again whether fd is within the fdt
+        * max_fds range, because we might be seeing a different
+        * files_fdtable() than iterate_fd(), assuming only RCU is
+        * protecting the read. In reality, iterate_fd() holds
+        * file_lock, which should ensure the fdt does not change while
+        * the lock is taken, but we are not aware whether this is
+        * guaranteed or not, so play safe.
+        */
+       if (fd < fdt->max_fds && lttng_close_on_exec(fd, fdt))
                flags |= O_CLOEXEC;
        if (IS_ERR(s)) {
                struct dentry *dentry = file->f_path.dentry;
@@ -248,10 +262,15 @@ void lttng_enumerate_task_fd(struct lttng_session *session,
                struct task_struct *p, char *tmp)
 {
        struct lttng_fd_ctx ctx = { .page = tmp, .session = session, .p = p };
+       struct files_struct *files;
 
        task_lock(p);
-       ctx.fdt = files_fdtable(p->files);
-       lttng_iterate_fd(p->files, 0, lttng_dump_one_fd, &ctx);
+       files = p->files;
+       if (!files)
+               goto end;
+       ctx.files = files;
+       lttng_iterate_fd(files, 0, lttng_dump_one_fd, &ctx);
+end:
        task_unlock(p);
 }
 
@@ -298,7 +317,7 @@ void lttng_enumerate_task_vm_maps(struct lttng_session *session,
                down_read(&mm->mmap_sem);
                while (map) {
                        if (map->vm_file)
-                               ino = map->vm_file->f_dentry->d_inode->i_ino;
+                               ino = map->vm_file->lttng_f_dentry->d_inode->i_ino;
                        else
                                ino = 0;
                        trace_lttng_statedump_vm_map(session, p, map, ino);
@@ -362,6 +381,9 @@ int lttng_list_interrupts(struct lttng_session *session)
 }
 #endif
 
+/*
+ * Called with task lock held.
+ */
 static
 void lttng_statedump_process_ns(struct lttng_session *session,
                struct task_struct *p,
@@ -373,8 +395,21 @@ void lttng_statedump_process_ns(struct lttng_session *session,
        struct nsproxy *proxy;
        struct pid_namespace *pid_ns;
 
+       /*
+        * Back and forth on locking strategy within Linux upstream for nsproxy.
+        * See Linux upstream commit 728dba3a39c66b3d8ac889ddbe38b5b1c264aec3
+        * "namespaces: Use task_lock and not rcu to protect nsproxy"
+        * for details.
+        */
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0) || \
+               LTTNG_UBUNTU_KERNEL_RANGE(3,13,11,36, 3,14,0,0) || \
+               LTTNG_UBUNTU_KERNEL_RANGE(3,16,1,11, 3,17,0,0) || \
+               LTTNG_RHEL_KERNEL_RANGE(3,10,0,229,13,0, 3,11,0,0,0,0))
+       proxy = p->nsproxy;
+#else
        rcu_read_lock();
        proxy = task_nsproxy(p);
+#endif
        if (proxy) {
                pid_ns = lttng_get_proxy_pid_ns(proxy);
                do {
@@ -386,7 +421,14 @@ void lttng_statedump_process_ns(struct lttng_session *session,
                trace_lttng_statedump_process_state(session,
                        p, type, mode, submode, status, NULL);
        }
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0) || \
+               LTTNG_UBUNTU_KERNEL_RANGE(3,13,11,36, 3,14,0,0) || \
+               LTTNG_UBUNTU_KERNEL_RANGE(3,16,1,11, 3,17,0,0) || \
+               LTTNG_RHEL_KERNEL_RANGE(3,10,0,229,13,0, 3,11,0,0,0,0))
+       /* (nothing) */
+#else
        rcu_read_unlock();
+#endif
 }
 
 static
@@ -482,6 +524,8 @@ int do_lttng_statedump(struct lttng_session *session)
                return ret;
        ret = lttng_enumerate_block_devices(session);
        switch (ret) {
+       case 0:
+               break;
        case -ENOSYS:
                printk(KERN_WARNING "LTTng: block device enumeration is not supported by kernel\n");
                break;