net: ipa: fix page free in two spots
v2 cover net: ipa: fix page free in two spots
When a receive buffer is not wrapped in an SKB and passed to the network stack, the (compound) page gets freed within the IPA driver. This is currently quite rare.
The pages are freed using __free_pages(), but they should instead be freed using page_put(). This series fixes this, in two spots.
These patches work for the current linus/master branch, but won't apply cleanly to earlier stable branches. (Nevertheless, the fix is a trivial substitution everwhere __free_pages() is called.)
Version 2 is just rebased on today's net/master branch.
-Alex
Alex Elder (2):
net: ipa: fix page free in ipa_endpoint_trans_release()
net: ipa: fix page free in ipa_endpoint_replenish_one()
drivers/net/ipa/ipa_endpoint.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-)
Signature · 1 line
2.32.0
PATCH v2 1/2 net: ipa: fix page free in ipa_endpoint_trans_release()
Currently the (possibly compound) page used for receive buffers are freed using __free_pages(). But according to this comment above the definition of that function, that's wrong:
If you want to use the page's reference count to decide when to free the allocation, you should allocate a compound page, and use put_page() instead of __free_pages().
Convert the call to __free_pages() in ipa_endpoint_trans_release() to use put_page() instead.
--- drivers/net/ipa/ipa_endpoint.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-)
drivers/net/ipa/ipa_endpoint.c +2 -5
@@ -1418,11 +1418,8 @@ void ipa_endpoint_trans_release(struct ipa_endpoint *endpoint, }else {struct page *page = trans->data; -if (page) { -u32 buffer_size = endpoint->config.rx.buffer_size; - -__free_pages (page,get_order (buffer_size)); - } +if (page) +put_page (page); } }
Signature · 1 line
2.32.0
PATCH v2 2/2 net: ipa: fix page free in ipa_endpoint_replenish_one()
In reply to Alex ElderCurrently the (possibly compound) pages used for receive buffers are freed using __free_pages(). But according to this comment above the definition of that function, that's wrong:
If you want to use the page's reference count to decide when to free the allocation, you should allocate a compound page, and use put_page() instead of __free_pages().
Convert the call to __free_pages() in ipa_endpoint_replenish_one() to use put_page() instead.
--- drivers/net/ipa/ipa_endpoint.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ipa/ipa_endpoint.c +1 -1
@@ -1095,7 +1095,7 @@ static int ipa_endpoint_replenish_one(struct ipa_endpoint *endpoint, ret =gsi_trans_page_add (trans, page, len, offset);if (ret) -__free_pages (page,get_order (buffer_size)); +put_page (page);else trans->data = page;/* transaction owns page now */
Signature · 1 line
2.32.0
This series was applied to netdev/net.git (master) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net.git (master) by Jakub Kicinski <[email protected]>:
Quoted 8 lines · On Thu, 26 May 2022 10:23:12 -0500 you wrote:
When a receive buffer is not wrapped in an SKB and passed to the network stack, the (compound) page gets freed within the IPA driver. This is currently quite rare.
The pages are freed using
__free_pages(), but they should instead be freed usingpage_put(). This series fixes this, in two spots.[...]
Here is the summary with links:
- [net,v2,1/2] net: ipa: fix page free in ipa_endpoint_trans_release()
https://git.kernel.org/netdev/net/c/155c0c90bca9 - [net,v2,2/2] net: ipa: fix page free in ipa_endpoint_replenish_one() https://git.kernel.org/netdev/net/c/70132763d5d2
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html