public inbox for ntb@lists.linux.dev
 help / color / mirror / Atom feed
From: Baruch Siach <baruch@tkos.co.il>
To: Koichiro Den <den@valinux.co.jp>
Cc: Jon Mason <jdmason@kudzu.us>,  Dave Jiang <dave.jiang@intel.com>,
	 Allen Hubbe <allenbh@gmail.com>,
	 ntb@lists.linux.dev
Subject: Re: [PATCH] NTB: ntb_transport: fix use after free
Date: Tue, 04 Nov 2025 07:42:53 +0200	[thread overview]
Message-ID: <87v7jqs68y.fsf@tarshish> (raw)
In-Reply-To: <y23vpg5ynutkwck3w6wynsy47yovpqywil2jqk4omg63qdocjb@xkcvg6h3jy2d> (Koichiro Den's message of "Tue, 4 Nov 2025 09:32:59 +0900")

Hi Koichiro,

Thanks for taking the time to review this.

On Tue, Nov 04 2025, Koichiro Den wrote:
> On Mon, Nov 03, 2025 at 07:23:19PM +0200, Baruch Siach wrote:
>> Don't call dmaengine_unmap_put() twice for the same pointer. This
>> results in mempool_free() being called on a freed element.
>
> While the second dmaengine_unmap_put() looks redundant at first glance, my
> understanding was that it compensates for the refcount decrement that would
> normally happen on the dmaengine completion path, which never occurs when
> dmaengine_submit() fails. Am I missing something?

I guess you are right. Now I see that dma_set_unmap() increments the
refcount. But I can't find the second refcount decrement in completion
path. I only see one call to dmaengine_unmap_put(). Can you enlighten
me?

Thanks,
baruch

>> Fixes: 6f57fd0578df ("NTB: convert to dmaengine_unmap_data")
>> Signed-off-by: Baruch Siach <baruch@tkos.co.il>
>> ---
>>  drivers/ntb/ntb_transport.c | 22 +++++++++-------------
>>  1 file changed, 9 insertions(+), 13 deletions(-)
>> 
>> diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c
>> index eb875e3db2e3..809fb09658b4 100644
>> --- a/drivers/ntb/ntb_transport.c
>> +++ b/drivers/ntb/ntb_transport.c
>> @@ -1573,14 +1573,14 @@ static int ntb_async_rx_submit(struct ntb_queue_entry *entry, void *offset)
>>  	unmap->addr[0] = dma_map_page(device->dev, virt_to_page(offset),
>>  				      pay_off, len, DMA_TO_DEVICE);
>>  	if (dma_mapping_error(device->dev, unmap->addr[0]))
>> -		goto err_get_unmap;
>> +		goto err_unmap;
>>  
>>  	unmap->to_cnt = 1;
>>  
>>  	unmap->addr[1] = dma_map_page(device->dev, virt_to_page(buf),
>>  				      buff_off, len, DMA_FROM_DEVICE);
>>  	if (dma_mapping_error(device->dev, unmap->addr[1]))
>> -		goto err_get_unmap;
>> +		goto err_unmap;
>>  
>>  	unmap->from_cnt = 1;
>>  
>> @@ -1588,7 +1588,7 @@ static int ntb_async_rx_submit(struct ntb_queue_entry *entry, void *offset)
>>  					     unmap->addr[0], len,
>>  					     DMA_PREP_INTERRUPT);
>>  	if (!txd)
>> -		goto err_get_unmap;
>> +		goto err_unmap;
>>  
>>  	txd->callback_result = ntb_rx_copy_callback;
>>  	txd->callback_param = entry;
>> @@ -1596,7 +1596,7 @@ static int ntb_async_rx_submit(struct ntb_queue_entry *entry, void *offset)
>>  
>>  	cookie = dmaengine_submit(txd);
>>  	if (dma_submit_error(cookie))
>> -		goto err_set_unmap;
>> +		goto err_unmap;
>>  
>>  	dmaengine_unmap_put(unmap);
>>  
>> @@ -1606,9 +1606,7 @@ static int ntb_async_rx_submit(struct ntb_queue_entry *entry, void *offset)
>>  
>>  	return 0;
>>  
>> -err_set_unmap:
>> -	dmaengine_unmap_put(unmap);
>> -err_get_unmap:
>> +err_unmap:
>>  	dmaengine_unmap_put(unmap);
>>  err:
>>  	return -ENXIO;
>> @@ -1854,14 +1852,14 @@ static int ntb_async_tx_submit(struct ntb_transport_qp *qp,
>>  	unmap->addr[0] = dma_map_page(device->dev, virt_to_page(buf),
>>  				      buff_off, len, DMA_TO_DEVICE);
>>  	if (dma_mapping_error(device->dev, unmap->addr[0]))
>> -		goto err_get_unmap;
>> +		goto err_unmap;
>>  
>>  	unmap->to_cnt = 1;
>>  
>>  	txd = device->device_prep_dma_memcpy(chan, dest, unmap->addr[0], len,
>>  					     DMA_PREP_INTERRUPT);
>>  	if (!txd)
>> -		goto err_get_unmap;
>> +		goto err_unmap;
>>  
>>  	txd->callback_result = ntb_tx_copy_callback;
>>  	txd->callback_param = entry;
>> @@ -1869,16 +1867,14 @@ static int ntb_async_tx_submit(struct ntb_transport_qp *qp,
>>  
>>  	cookie = dmaengine_submit(txd);
>>  	if (dma_submit_error(cookie))
>> -		goto err_set_unmap;
>> +		goto err_unmap;
>>  
>>  	dmaengine_unmap_put(unmap);
>>  
>>  	dma_async_issue_pending(chan);
>>  
>>  	return 0;
>> -err_set_unmap:
>> -	dmaengine_unmap_put(unmap);
>> -err_get_unmap:
>> +err_unmap:
>>  	dmaengine_unmap_put(unmap);
>>  err:
>>  	return -ENXIO;
>> -- 
>> 2.51.0
>> 

-- 
                                                     ~. .~   Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
   - baruch@tkos.co.il - tel: +972.52.368.4656, http://www.tkos.co.il -

  reply	other threads:[~2025-11-04  5:42 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-11-03 17:23 [PATCH] NTB: ntb_transport: fix use after free Baruch Siach
2025-11-03 17:45 ` Dave Jiang
2025-11-04  0:32 ` Koichiro Den
2025-11-04  5:42   ` Baruch Siach [this message]
2025-11-06  3:59     ` Koichiro Den
2025-11-06  6:02       ` Baruch Siach

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=87v7jqs68y.fsf@tarshish \
    --to=baruch@tkos.co.il \
    --cc=allenbh@gmail.com \
    --cc=dave.jiang@intel.com \
    --cc=den@valinux.co.jp \
    --cc=jdmason@kudzu.us \
    --cc=ntb@lists.linux.dev \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox