[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <afc69260-b07d-107a-20e2-132c171ce85f@huawei.com>
Date: Tue, 8 May 2018 09:56:37 +0800
From: Chao Yu <yuchao0@...wei.com>
To: Jaegeuk Kim <jaegeuk@...nel.org>
CC: <linux-f2fs-devel@...ts.sourceforge.net>,
<linux-kernel@...r.kernel.org>, <chao@...nel.org>
Subject: Re: [PATCH v4] f2fs: fix to avoid race during access gc_thread
pointer
On 2018/5/8 5:36, Jaegeuk Kim wrote:
> On 05/07, Chao Yu wrote:
>> Thread A Thread B Thread C
>> - f2fs_remount
>> - stop_gc_thread
>> - f2fs_sbi_store
>> - issue_discard_thread
>> sbi->gc_thread = NULL;
>> sbi->gc_thread->gc_wake = 1
>> access sbi->gc_thread->gc_urgent
>>
>> Previously, we allocate memory for sbi->gc_thread based on background
>> gc thread mount option, the memory can be released if we turn off
>> that mount option, but still there are several places access gc_thread
>> pointer without considering race condition, result in NULL point
>> dereference.
>>
>> In order to fix this issue, introduce gc_rwsem to exclude those operations.
>>
>> Signed-off-by: Chao Yu <yuchao0@...wei.com>
>> ---
>> v4:
>> - use introduced sbi.gc_rwsem lock instead of sb.s_umount.
>
> We can use this first.
>
>>>From e62e8d3ece6ee8a4aeac8ffd6161d25851f8b3f0 Mon Sep 17 00:00:00 2001
> From: Jaegeuk Kim <jaegeuk@...nel.org>
> Date: Mon, 7 May 2018 14:22:40 -0700
> Subject: [PATCH] f2fs: introduce sbi->gc_mode to determine the policy
>
> This is to avoid sbi->gc_thread pointer access.
>
> Signed-off-by: Jaegeuk Kim <jaegeuk@...nel.org>
> ---
> fs/f2fs/f2fs.h | 8 ++++++++
> fs/f2fs/gc.c | 28 ++++++++++++----------------
> fs/f2fs/gc.h | 2 --
> fs/f2fs/segment.c | 4 ++--
> fs/f2fs/sysfs.c | 33 +++++++++++++++++++++++++--------
> 5 files changed, 47 insertions(+), 28 deletions(-)
>
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index 80490a7991a7..779d8b26878c 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -1065,6 +1065,13 @@ enum {
> MAX_TIME,
> };
>
> +enum {
> + GC_NORMAL,
> + GC_IDLE_CB,
> + GC_IDLE_GREEDY,
> + GC_URGENT,
> +};
> +
> enum {
> WHINT_MODE_OFF, /* not pass down write hints */
> WHINT_MODE_USER, /* try to pass down hints given by users */
> @@ -1193,6 +1200,7 @@ struct f2fs_sb_info {
> struct mutex gc_mutex; /* mutex for GC */
> struct f2fs_gc_kthread *gc_thread; /* GC thread */
> unsigned int cur_victim_sec; /* current victim section num */
> + unsigned int gc_mode; /* current GC state */
>
> /* threshold for gc trials on pinned files */
> u64 gc_pin_file_threshold;
> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> index 9bb2ddbbed1e..7ec8ea75dfde 100644
> --- a/fs/f2fs/gc.c
> +++ b/fs/f2fs/gc.c
> @@ -76,7 +76,7 @@ static int gc_thread_func(void *data)
> * invalidated soon after by user update or deletion.
> * So, I'd like to wait some time to collect dirty segments.
> */
> - if (gc_th->gc_urgent) {
> + if (sbi->gc_mode == GC_URGENT) {
> wait_ms = gc_th->urgent_sleep_time;
> mutex_lock(&sbi->gc_mutex);
> goto do_gc;
> @@ -131,8 +131,6 @@ int start_gc_thread(struct f2fs_sb_info *sbi)
> gc_th->max_sleep_time = DEF_GC_THREAD_MAX_SLEEP_TIME;
> gc_th->no_gc_sleep_time = DEF_GC_THREAD_NOGC_SLEEP_TIME;
>
> - gc_th->gc_idle = 0;
> - gc_th->gc_urgent = 0;
> gc_th->gc_wake= 0;
>
> sbi->gc_thread = gc_th;
> @@ -158,21 +156,19 @@ void stop_gc_thread(struct f2fs_sb_info *sbi)
> sbi->gc_thread = NULL;
> }
>
> -static int select_gc_type(struct f2fs_gc_kthread *gc_th, int gc_type)
> +static int select_gc_type(struct f2fs_sb_info *sbi, int gc_type)
> {
> int gc_mode = (gc_type == BG_GC) ? GC_CB : GC_GREEDY;
>
> - if (!gc_th)
> - return gc_mode;
> -
> - if (gc_th->gc_idle) {
> - if (gc_th->gc_idle == 1)
> - gc_mode = GC_CB;
> - else if (gc_th->gc_idle == 2)
> - gc_mode = GC_GREEDY;
> - }
> - if (gc_th->gc_urgent)
> + switch (sbi->gc_mode) {
> + case GC_IDLE_CB:
> + gc_mode = GC_CB;
> + break;
> + case GC_IDLE_GREEDY:
> + case GC_URGENT:
> gc_mode = GC_GREEDY;
> + break;
> + }
> return gc_mode;
> }
>
> @@ -187,7 +183,7 @@ static void select_policy(struct f2fs_sb_info *sbi, int gc_type,
> p->max_search = dirty_i->nr_dirty[type];
> p->ofs_unit = 1;
> } else {
> - p->gc_mode = select_gc_type(sbi->gc_thread, gc_type);
> + p->gc_mode = select_gc_type(sbi, gc_type);
> p->dirty_segmap = dirty_i->dirty_segmap[DIRTY];
> p->max_search = dirty_i->nr_dirty[DIRTY];
> p->ofs_unit = sbi->segs_per_sec;
> @@ -195,7 +191,7 @@ static void select_policy(struct f2fs_sb_info *sbi, int gc_type,
>
> /* we need to check every dirty segments in the FG_GC case */
> if (gc_type != FG_GC &&
> - (sbi->gc_thread && !sbi->gc_thread->gc_urgent) &&
> + (sbi->gc_mode != GC_URGENT) &&
> p->max_search > sbi->max_victim_search)
> p->max_search = sbi->max_victim_search;
>
> diff --git a/fs/f2fs/gc.h b/fs/f2fs/gc.h
> index b0045d4c8d1e..c8619e408009 100644
> --- a/fs/f2fs/gc.h
> +++ b/fs/f2fs/gc.h
> @@ -36,8 +36,6 @@ struct f2fs_gc_kthread {
> unsigned int no_gc_sleep_time;
>
> /* for changing gc mode */
> - unsigned int gc_idle;
> - unsigned int gc_urgent;
> unsigned int gc_wake;
> };
>
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index 503a98abdb2f..5a7ed06291e0 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -177,7 +177,7 @@ bool need_SSR(struct f2fs_sb_info *sbi)
>
> if (test_opt(sbi, LFS))
> return false;
> - if (sbi->gc_thread && sbi->gc_thread->gc_urgent)
> + if (sbi->gc_mode == GC_URGENT)
> return true;
>
> return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs +
> @@ -1418,7 +1418,7 @@ static int issue_discard_thread(void *data)
> if (dcc->discard_wake)
> dcc->discard_wake = 0;
>
> - if (sbi->gc_thread && sbi->gc_thread->gc_urgent)
> + if (sbi->gc_mode == GC_URGENT)
> init_discard_policy(&dpolicy, DPOLICY_FORCE, 1);
>
> sb_start_intwrite(sbi->sb);
> diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c
> index 6d8d8f41e517..dd940d156af6 100644
> --- a/fs/f2fs/sysfs.c
> +++ b/fs/f2fs/sysfs.c
> @@ -248,16 +248,33 @@ static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
> if (!strcmp(a->attr.name, "trim_sections"))
> return -EINVAL;
>
> + if (!strcmp(a->attr.name, "gc_urgent")) {
> + if (t >= 1) {
> + sbi->gc_mode = GC_URGENT;
> + if (sbi->gc_thread) {
> + wake_up_interruptible_all(
> + &sbi->gc_thread->gc_wait_queue_head);
> + wake_up_discard_thread(sbi, true);
> + }
> + } else {
> + sbi->gc_mode = GC_NORMAL;
> + }
> + return count;
> + }
> + if (!strcmp(a->attr.name, "gc_idle")) {
> + if (t == GC_IDLE_CB)
> + sbi->gc_mode = GC_IDLE_CB;
> + else if (t == GC_IDLE_GREEDY)
> + sbi->gc_mode = GC_IDLE_GREEDY;
> + else
> + sbi->gc_mode = GC_NORMAL;
> + return count;
> + }
> +
> *ui = t;
>
> if (!strcmp(a->attr.name, "iostat_enable") && *ui == 0)
> f2fs_reset_iostat(sbi);
> - if (!strcmp(a->attr.name, "gc_urgent") && t == 1 && sbi->gc_thread) {
> - sbi->gc_thread->gc_wake = 1;
> - wake_up_interruptible_all(&sbi->gc_thread->gc_wait_queue_head);
> - wake_up_discard_thread(sbi, true);
> - }
> -
> return count;
> }
>
> @@ -349,8 +366,8 @@ F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_urgent_sleep_time,
> F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
> F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
> F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
All above parameters will be accessed via '*ui = t;', in order to avoid that, we
need to move all gc_* configuration fields into sbi, so leaving f2fs_gc_kthread
all most empty. IMO, just merge my v1 patch is OK? since that patch makes better
encapsulation, so that background GC related parameter can still locate in
independent f2fs_gc_kthread structure rather in common f2fs_sb_info structure.
Thanks,
> -F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
> -F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_urgent, gc_urgent);
> +F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_idle, gc_mode);
> +F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_urgent, gc_mode);
> F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
> F2FS_RW_ATTR(DCC_INFO, discard_cmd_control, max_small_discards, max_discards);
> F2FS_RW_ATTR(DCC_INFO, discard_cmd_control, discard_granularity, discard_granularity);
>
Powered by blists - more mailing lists