Linux-Trace-Devel Archive on lore.kernel.org
 help / color / Atom feed
From: "Yordan Karadzhov (VMware)" <y.karadz@gmail.com>
To: Steven Rostedt <rostedt@goodmis.org>
Cc: linux-trace-devel@vger.kernel.org
Subject: Re: [PATCH v2 10/20] kernel-shark: Start using data streams
Date: Fri, 6 Nov 2020 16:31:58 +0200
Message-ID: <d88af9bb-1f15-04cf-230c-ea59b3239931@gmail.com> (raw)
In-Reply-To: <20201105131735.3d559bce@gandalf.local.home>



On 5.11.20 г. 20:17 ч., Steven Rostedt wrote:
> On Thu, 5 Nov 2020 16:58:51 +0200
> "Yordan Karadzhov (VMware)" <y.karadz@gmail.com> wrote:
> 
>> I see the problem. This is definitely wrong.
>>
>> What if in addition to "n_streams" I add another counter called
>> "last_stream_added" and initialize this counter to -1?
>>
>> Then I can add streams like this:
>>
>> 	kshark_ctx->stream[++kshark_ctx->last_stream_added] = stream;
>> 	++kshark_ctx->n_streams;
>>
>>> You may want to do instead:
> 
> I'm thinking of doing something like this:
> 
> struct kshark_ctx {
> 	[..]
> 	unsigned int free_stream;
> 	[..]
> };
> 
> 
> 	if (kshark_ctx->free_stream >= kshark_ctx->n_streams) {
> 		kshark_ctx->stream[++kshark_ctx->n_streams] = stream;
> 		kshark_ctx->free_stream = kshark_ctx->n_streams;
> 
> /* BTW, the stream array should be allocated to the n_streams, and
>     reallocated when it grows.  I don't think we want a huge stream array to
>     handle all the bits when not used. */
> 
> 	} else {
> 		int new_stream = kshark_ctx->free_stream;
> 
> 		kshark_ctx->free_stream = kshark_index(kshark_ctx->stream[new_stream]);
> 		kshark_ctx->stream[new_stream] = stream;
> 	}
> 
> 
> For freeing (index i):
> 
> 	kshark_ctx->stream[i] = kshark_ptr(kshark_ctx->free_stream);
> 	kshark_ctx->free_stream = i;
> 
> 
> We could define the following (note, I just used these names for the
> functions, they could be named something else):
> 
> 
> #define KSHARK_INDEX_MASK ((1 << NR_OF_BITS_FOR_STREAM) - 1)
> 
> #define KSHARK_INVALID_STREAM (~((1UL << NR_OF_BITS_FOR_STREAM) - 1))
> 
> static inline int kshark_index(void *ptr)
> {
> 	unsigned long index = (unsigned long)ptr;
> 
> 	return (int)(index & KSHARK_INDEX_MASK);
> }
> 
> static inline void *kshark_ptr(unsigned int index)
> {
> 	unsigned long p;
> 
> 	p = KSHARK_INVALID_STREAM | index;
> 
> 	return (void *)p;
> }
> 
> The KSHARK_INVALID_STREAM and KSHARK_INDEX_MASK, would allow us to do
> something like this if we wanted to loop through all streams:
> 
> static inline bool kshark_is_valid_stream(void *ptr)
> {
> 	unsigned long p = (unsigned long)ptr;
> 
> 	return (p & KSHARK_INVALID_STREAM) == KSHARK_INVALID_STREAM;
> }
> 
> The above works because the address of setting all those bits, would put
> the address into the kernel space (illegal user space address).
> 
> 
> 	for (i = 0; i < kshark_ctx->n_streams; i++) {
> 		if (!kshark_is_valid_stream(kshark_ctx->stream[i]))
> 			continue;
> 		/* process valid stream */
> 	}
>

Hi Steven,

I am not sure I understand correctly your pseudo-code, so please correct 
me if my interpretation is wrong.

In the normal case when a new stream is added the corresponding object 
will be allocated and added to the array of pointers. Later if a stream 
is removed, instead of freeing the memory we will just manipulate it 
pointer so that it point to nowhere and this manipulation can be 
detected by the kshark_is_valid_stream(). Now if we want to add stream 
again, we will take the broken pointer, will restore its original value 
and will reuse the object without a new allocations.

And at the very end we will have to free all pointers (original or 
manipulated).

Is this what you are suggesting?

Thanks!
Y.


> -- Steve
> 

  reply index

Thread overview: 47+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-12 13:35 [PATCH v2 00/20] Start KernelShark v2 transformation Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 01/20] kernel-shark: Start introducing KernelShark 2.0 Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 02/20] kernel-shark: Use only signed types in kshark_entry Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 03/20] kernel-shark: Introduce libkshark-hash Yordan Karadzhov (VMware)
2020-10-12 14:05   ` Steven Rostedt
2020-10-12 14:05     ` Steven Rostedt
2020-10-12 14:18       ` Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 04/20] kernel-shark: Introduce Data streams Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 05/20] kernel-shark: Add stream_id to kshark_entry Yordan Karadzhov (VMware)
2020-10-13  0:05   ` Steven Rostedt
2020-10-29 10:08     ` Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 06/20] kernel-shark: Rename static methods in libkshark Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 07/20] kernel-shark: Add basic methods for Data streams Yordan Karadzhov (VMware)
2020-10-13  0:18   ` Steven Rostedt
2020-10-29 10:10     ` Yordan Karadzhov (VMware)
2020-10-29 14:04       ` Steven Rostedt
2020-10-29 14:49         ` Yordan Karadzhov (VMware)
2020-10-30  1:57           ` Steven Rostedt
2020-11-03 13:38             ` Yordan Karadzhov (VMware)
2020-11-04 15:41               ` Steven Rostedt
2020-11-05 14:35                 ` Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 08/20] kernel-shark: Housekeeping before implementing stream interface Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 09/20] kernel-shark: Add stream interface for trace-cmd data Yordan Karadzhov (VMware)
2020-10-13  0:44   ` Steven Rostedt
2020-10-29 11:16     ` Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 10/20] kernel-shark: Start using data streams Yordan Karadzhov (VMware)
2020-10-14 18:56   ` Steven Rostedt
2020-11-05 14:58     ` Yordan Karadzhov (VMware)
2020-11-05 18:17       ` Steven Rostedt
2020-11-06 14:31         ` Yordan Karadzhov (VMware) [this message]
2020-11-06 15:18           ` Steven Rostedt
2020-11-09 14:49             ` Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 11/20] kernel-shark: Remove dead code Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 12/20] kernel-shark: Redesign the plugin interface Yordan Karadzhov (VMware)
2020-10-14 21:09   ` Steven Rostedt
2020-10-12 13:35 ` [PATCH v2 13/20] kernel-shark: Complete the stream integration Yordan Karadzhov (VMware)
2020-10-14 23:52   ` Steven Rostedt
2020-10-12 13:35 ` [PATCH v2 14/20] kernel-shark: Provide merging of multiple data streams Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 15/20] kernel-shark: Integrate the stream definitions with data model Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 16/20] kernel-shark: Use only signed types for model defs Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 17/20] kernel-shark: Add ksmodel_get_bin() Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 18/20] kernel-shark: Protect ksmodel_set_in_range_bining() Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 19/20] kernel-shark: Add methods for time calibration Yordan Karadzhov (VMware)
2020-10-12 13:35 ` [PATCH v2 20/20] kernel-shark: Integrate streams with libkshark-configio Yordan Karadzhov (VMware)
2020-11-05 19:22   ` Steven Rostedt
2020-11-09 14:55     ` Yordan Karadzhov (VMware)
2020-11-09 15:28       ` Steven Rostedt

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=d88af9bb-1f15-04cf-230c-ea59b3239931@gmail.com \
    --to=y.karadz@gmail.com \
    --cc=linux-trace-devel@vger.kernel.org \
    --cc=rostedt@goodmis.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-Trace-Devel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-trace-devel/0 linux-trace-devel/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-trace-devel linux-trace-devel/ https://lore.kernel.org/linux-trace-devel \
		linux-trace-devel@vger.kernel.org
	public-inbox-index linux-trace-devel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-trace-devel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git