* [PATCH] io_uring: add more uring info to fdinfo for debug
@ 2021-09-13 13:08 Hao Xu
2021-09-15 13:01 ` Hao Xu
2021-09-15 15:26 ` Jens Axboe
0 siblings, 2 replies; 6+ messages in thread
From: Hao Xu @ 2021-09-13 13:08 UTC (permalink / raw)
To: Jens Axboe; +Cc: io-uring, Pavel Begunkov, Joseph Qi
Developers may need some uring info to help themselves debug and address
issues, these info includes sqring/cqring head/tail and the detail
sqe/cqe info, which is very useful when it stucks.
Signed-off-by: Hao Xu <[email protected]>
---
fs/io_uring.c | 59 +++++++++++++++++++++++++++++++++++++++++++++++----
1 file changed, 55 insertions(+), 4 deletions(-)
diff --git a/fs/io_uring.c b/fs/io_uring.c
index f795ad281038..ac048592a3e8 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -9950,8 +9950,48 @@ static int io_uring_show_cred(struct seq_file *m, unsigned int id,
static void __io_uring_show_fdinfo(struct io_ring_ctx *ctx, struct seq_file *m)
{
struct io_sq_data *sq = NULL;
+ struct io_overflow_cqe *ocqe;
+ struct io_rings *r = ctx->rings;
+ unsigned int sq_mask = ctx->sq_entries - 1, cq_mask = ctx->cq_entries - 1;
+ unsigned int cached_sq_head = ctx->cached_sq_head;
+ unsigned int cached_cq_tail = ctx->cached_cq_tail;
+ unsigned int sq_head = READ_ONCE(r->sq.head);
+ unsigned int sq_tail = READ_ONCE(r->sq.tail);
+ unsigned int cq_head = READ_ONCE(r->cq.head);
+ unsigned int cq_tail = READ_ONCE(r->cq.tail);
bool has_lock;
- int i;
+ unsigned int i;
+
+ /*
+ * we may get imprecise sqe and cqe info if uring is actively running
+ * since we get cached_sq_head and cached_cq_tail without uring_lock
+ * and sq_tail and cq_head are changed by userspace. But it's ok since
+ * we usually use these info when it is stuck.
+ */
+ seq_printf(m, "SqHead:\t%u\n", sq_head & sq_mask);
+ seq_printf(m, "SqTail:\t%u\n", sq_tail & sq_mask);
+ seq_printf(m, "CachedSqHead:\t%u\n", cached_sq_head & sq_mask);
+ seq_printf(m, "CqHead:\t%u\n", cq_head & cq_mask);
+ seq_printf(m, "CqTail:\t%u\n", cq_tail & cq_mask);
+ seq_printf(m, "CachedCqTail:\t%u\n", cached_cq_tail & cq_mask);
+ seq_printf(m, "SQEs:\t%u\n", sq_tail - cached_sq_head);
+ for (i = cached_sq_head; i < sq_tail; i++) {
+ unsigned int sq_idx = READ_ONCE(ctx->sq_array[i & sq_mask]);
+
+ if (likely(sq_idx <= sq_mask)) {
+ struct io_uring_sqe *sqe = &ctx->sq_sqes[sq_idx];
+
+ seq_printf(m, "%5u: opcode:%d, fd:%d, flags:%x, user_data:%llu\n",
+ sq_idx, sqe->opcode, sqe->fd, sqe->flags, sqe->user_data);
+ }
+ }
+ seq_printf(m, "CQEs:\t%u\n", cached_cq_tail - cq_head);
+ for (i = cq_head; i < cached_cq_tail; i++) {
+ struct io_uring_cqe *cqe = &r->cqes[i & cq_mask];
+
+ seq_printf(m, "%5u: user_data:%llu, res:%d, flag:%x\n",
+ i & cq_mask, cqe->user_data, cqe->res, cqe->flags);
+ }
/*
* Avoid ABBA deadlock between the seq lock and the io_uring mutex,
@@ -9993,7 +10033,10 @@ static void __io_uring_show_fdinfo(struct io_ring_ctx *ctx, struct seq_file *m)
xa_for_each(&ctx->personalities, index, cred)
io_uring_show_cred(m, index, cred);
}
- seq_printf(m, "PollList:\n");
+ if (has_lock)
+ mutex_unlock(&ctx->uring_lock);
+
+ seq_puts(m, "PollList:\n");
spin_lock(&ctx->completion_lock);
for (i = 0; i < (1U << ctx->cancel_hash_bits); i++) {
struct hlist_head *list = &ctx->cancel_hash[i];
@@ -10003,9 +10046,17 @@ static void __io_uring_show_fdinfo(struct io_ring_ctx *ctx, struct seq_file *m)
seq_printf(m, " op=%d, task_works=%d\n", req->opcode,
req->task->task_works != NULL);
}
+
+ seq_puts(m, "CqOverflowList:\n");
+ list_for_each_entry(ocqe, &ctx->cq_overflow_list, list) {
+ struct io_uring_cqe *cqe = &ocqe->cqe;
+
+ seq_printf(m, " user_data=%llu, res=%d, flags=%x\n",
+ cqe->user_data, cqe->res, cqe->flags);
+
+ }
+
spin_unlock(&ctx->completion_lock);
- if (has_lock)
- mutex_unlock(&ctx->uring_lock);
}
static void io_uring_show_fdinfo(struct seq_file *m, struct file *f)
--
2.24.4
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH] io_uring: add more uring info to fdinfo for debug
2021-09-13 13:08 [PATCH] io_uring: add more uring info to fdinfo for debug Hao Xu
@ 2021-09-15 13:01 ` Hao Xu
2021-09-15 15:26 ` Jens Axboe
1 sibling, 0 replies; 6+ messages in thread
From: Hao Xu @ 2021-09-15 13:01 UTC (permalink / raw)
To: Jens Axboe; +Cc: io-uring, Pavel Begunkov, Joseph Qi
ping...
在 2021/9/13 下午9:08, Hao Xu 写道:
> Developers may need some uring info to help themselves debug and address
> issues, these info includes sqring/cqring head/tail and the detail
> sqe/cqe info, which is very useful when it stucks.
>
> Signed-off-by: Hao Xu <[email protected]>
> ---
> fs/io_uring.c | 59 +++++++++++++++++++++++++++++++++++++++++++++++----
> 1 file changed, 55 insertions(+), 4 deletions(-)
>
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index f795ad281038..ac048592a3e8 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -9950,8 +9950,48 @@ static int io_uring_show_cred(struct seq_file *m, unsigned int id,
> static void __io_uring_show_fdinfo(struct io_ring_ctx *ctx, struct seq_file *m)
> {
> struct io_sq_data *sq = NULL;
> + struct io_overflow_cqe *ocqe;
> + struct io_rings *r = ctx->rings;
> + unsigned int sq_mask = ctx->sq_entries - 1, cq_mask = ctx->cq_entries - 1;
> + unsigned int cached_sq_head = ctx->cached_sq_head;
> + unsigned int cached_cq_tail = ctx->cached_cq_tail;
> + unsigned int sq_head = READ_ONCE(r->sq.head);
> + unsigned int sq_tail = READ_ONCE(r->sq.tail);
> + unsigned int cq_head = READ_ONCE(r->cq.head);
> + unsigned int cq_tail = READ_ONCE(r->cq.tail);
> bool has_lock;
> - int i;
> + unsigned int i;
> +
> + /*
> + * we may get imprecise sqe and cqe info if uring is actively running
> + * since we get cached_sq_head and cached_cq_tail without uring_lock
> + * and sq_tail and cq_head are changed by userspace. But it's ok since
> + * we usually use these info when it is stuck.
> + */
> + seq_printf(m, "SqHead:\t%u\n", sq_head & sq_mask);
> + seq_printf(m, "SqTail:\t%u\n", sq_tail & sq_mask);
> + seq_printf(m, "CachedSqHead:\t%u\n", cached_sq_head & sq_mask);
> + seq_printf(m, "CqHead:\t%u\n", cq_head & cq_mask);
> + seq_printf(m, "CqTail:\t%u\n", cq_tail & cq_mask);
> + seq_printf(m, "CachedCqTail:\t%u\n", cached_cq_tail & cq_mask);
> + seq_printf(m, "SQEs:\t%u\n", sq_tail - cached_sq_head);
> + for (i = cached_sq_head; i < sq_tail; i++) {
> + unsigned int sq_idx = READ_ONCE(ctx->sq_array[i & sq_mask]);
> +
> + if (likely(sq_idx <= sq_mask)) {
> + struct io_uring_sqe *sqe = &ctx->sq_sqes[sq_idx];
> +
> + seq_printf(m, "%5u: opcode:%d, fd:%d, flags:%x, user_data:%llu\n",
> + sq_idx, sqe->opcode, sqe->fd, sqe->flags, sqe->user_data);
> + }
> + }
> + seq_printf(m, "CQEs:\t%u\n", cached_cq_tail - cq_head);
> + for (i = cq_head; i < cached_cq_tail; i++) {
> + struct io_uring_cqe *cqe = &r->cqes[i & cq_mask];
> +
> + seq_printf(m, "%5u: user_data:%llu, res:%d, flag:%x\n",
> + i & cq_mask, cqe->user_data, cqe->res, cqe->flags);
> + }
>
> /*
> * Avoid ABBA deadlock between the seq lock and the io_uring mutex,
> @@ -9993,7 +10033,10 @@ static void __io_uring_show_fdinfo(struct io_ring_ctx *ctx, struct seq_file *m)
> xa_for_each(&ctx->personalities, index, cred)
> io_uring_show_cred(m, index, cred);
> }
> - seq_printf(m, "PollList:\n");
> + if (has_lock)
> + mutex_unlock(&ctx->uring_lock);
> +
> + seq_puts(m, "PollList:\n");
> spin_lock(&ctx->completion_lock);
> for (i = 0; i < (1U << ctx->cancel_hash_bits); i++) {
> struct hlist_head *list = &ctx->cancel_hash[i];
> @@ -10003,9 +10046,17 @@ static void __io_uring_show_fdinfo(struct io_ring_ctx *ctx, struct seq_file *m)
> seq_printf(m, " op=%d, task_works=%d\n", req->opcode,
> req->task->task_works != NULL);
> }
> +
> + seq_puts(m, "CqOverflowList:\n");
> + list_for_each_entry(ocqe, &ctx->cq_overflow_list, list) {
> + struct io_uring_cqe *cqe = &ocqe->cqe;
> +
> + seq_printf(m, " user_data=%llu, res=%d, flags=%x\n",
> + cqe->user_data, cqe->res, cqe->flags);
> +
> + }
> +
> spin_unlock(&ctx->completion_lock);
> - if (has_lock)
> - mutex_unlock(&ctx->uring_lock);
> }
>
> static void io_uring_show_fdinfo(struct seq_file *m, struct file *f)
>
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] io_uring: add more uring info to fdinfo for debug
2021-09-13 13:08 [PATCH] io_uring: add more uring info to fdinfo for debug Hao Xu
2021-09-15 13:01 ` Hao Xu
@ 2021-09-15 15:26 ` Jens Axboe
2021-09-15 15:31 ` Pavel Begunkov
1 sibling, 1 reply; 6+ messages in thread
From: Jens Axboe @ 2021-09-15 15:26 UTC (permalink / raw)
To: Hao Xu; +Cc: io-uring, Pavel Begunkov, Joseph Qi
On 9/13/21 7:08 AM, Hao Xu wrote:
> Developers may need some uring info to help themselves debug and address
> issues, these info includes sqring/cqring head/tail and the detail
> sqe/cqe info, which is very useful when it stucks.
I think this is a good addition, more info to help you debug a stuck case
is always good. I'll queue this up for 5.16.
--
Jens Axboe
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] io_uring: add more uring info to fdinfo for debug
2021-09-15 15:26 ` Jens Axboe
@ 2021-09-15 15:31 ` Pavel Begunkov
2021-09-15 15:32 ` Jens Axboe
0 siblings, 1 reply; 6+ messages in thread
From: Pavel Begunkov @ 2021-09-15 15:31 UTC (permalink / raw)
To: Jens Axboe, Hao Xu; +Cc: io-uring, Joseph Qi
On 9/15/21 4:26 PM, Jens Axboe wrote:
> On 9/13/21 7:08 AM, Hao Xu wrote:
>> Developers may need some uring info to help themselves debug and address
>> issues, these info includes sqring/cqring head/tail and the detail
>> sqe/cqe info, which is very useful when it stucks.
>
> I think this is a good addition, more info to help you debug a stuck case
> is always good. I'll queue this up for 5.16.
Are there limits how much we can print? I remember people were couldn't
even show a list of CPUs (was it proc?). The overflow list may be huge.
--
Pavel Begunkov
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] io_uring: add more uring info to fdinfo for debug
2021-09-15 15:31 ` Pavel Begunkov
@ 2021-09-15 15:32 ` Jens Axboe
2021-09-16 6:29 ` Hao Xu
0 siblings, 1 reply; 6+ messages in thread
From: Jens Axboe @ 2021-09-15 15:32 UTC (permalink / raw)
To: Pavel Begunkov, Hao Xu; +Cc: io-uring, Joseph Qi
On 9/15/21 9:31 AM, Pavel Begunkov wrote:
> On 9/15/21 4:26 PM, Jens Axboe wrote:
>> On 9/13/21 7:08 AM, Hao Xu wrote:
>>> Developers may need some uring info to help themselves debug and address
>>> issues, these info includes sqring/cqring head/tail and the detail
>>> sqe/cqe info, which is very useful when it stucks.
>>
>> I think this is a good addition, more info to help you debug a stuck case
>> is always good. I'll queue this up for 5.16.
>
> Are there limits how much we can print? I remember people were couldn't
> even show a list of CPUs (was it proc?). The overflow list may be huge.
It's using seq_file, so I _think_ we should be fine here. Not sure when/if
it truncates.
--
Jens Axboe
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] io_uring: add more uring info to fdinfo for debug
2021-09-15 15:32 ` Jens Axboe
@ 2021-09-16 6:29 ` Hao Xu
0 siblings, 0 replies; 6+ messages in thread
From: Hao Xu @ 2021-09-16 6:29 UTC (permalink / raw)
To: Jens Axboe, Pavel Begunkov; +Cc: io-uring, Joseph Qi
在 2021/9/15 下午11:32, Jens Axboe 写道:
> On 9/15/21 9:31 AM, Pavel Begunkov wrote:
>> On 9/15/21 4:26 PM, Jens Axboe wrote:
>>> On 9/13/21 7:08 AM, Hao Xu wrote:
>>>> Developers may need some uring info to help themselves debug and address
>>>> issues, these info includes sqring/cqring head/tail and the detail
>>>> sqe/cqe info, which is very useful when it stucks.
>>>
>>> I think this is a good addition, more info to help you debug a stuck case
>>> is always good. I'll queue this up for 5.16.
>>
>> Are there limits how much we can print? I remember people were couldn't
>> even show a list of CPUs (was it proc?). The overflow list may be huge.
>
> It's using seq_file, so I _think_ we should be fine here. Not sure when/if
> it truncates.
>
It seems the default buf size of seq_file is PAGE_SIZE.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-09-16 6:29 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2021-09-13 13:08 [PATCH] io_uring: add more uring info to fdinfo for debug Hao Xu
2021-09-15 13:01 ` Hao Xu
2021-09-15 15:26 ` Jens Axboe
2021-09-15 15:31 ` Pavel Begunkov
2021-09-15 15:32 ` Jens Axboe
2021-09-16 6:29 ` Hao Xu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox