On 2020/8/18 16:08, Eric Dumazet wrote:
>
>
> On 8/16/20 12:08 AM, Coly Li wrote:
>> The original problem was from nvme-over-tcp code, who mistakenly uses
>> kernel_sendpage() to send pages allocated by __get_free_pages() without
>> __GFP_COMP flag. Such pages don't have refcount (page_count is 0
On 8/16/20 12:08 AM, Coly Li wrote:
> The original problem was from nvme-over-tcp code, who mistakenly uses
> kernel_sendpage() to send pages allocated by __get_free_pages() without
> __GFP_COMP flag. Such pages don't have refcount (page_count is 0) on
> tail pages, sending them by kernel_sendpa
On Mon, Aug 17, 2020 at 12:12:12PM -0700, Cong Wang wrote:
>
> So netdev people will have to understand and support PageSlab() or
> page_count()?
Yes. As they came up with that contrived rule what is acceptable
for sendpage. No one else really knows and other subsystems like the
block layer are
On Sun, Aug 16, 2020 at 10:45 PM Christoph Hellwig wrote:
>
> On Sun, Aug 16, 2020 at 10:55:09AM -0700, Cong Wang wrote:
> > On Sun, Aug 16, 2020 at 1:36 AM Coly Li wrote:
> > >
> > > The original problem was from nvme-over-tcp code, who mistakenly uses
> > > kernel_sendpage() to send pages alloc
On Sun, Aug 16, 2020 at 10:55:09AM -0700, Cong Wang wrote:
> On Sun, Aug 16, 2020 at 1:36 AM Coly Li wrote:
> >
> > The original problem was from nvme-over-tcp code, who mistakenly uses
> > kernel_sendpage() to send pages allocated by __get_free_pages() without
> > __GFP_COMP flag. Such pages don'
On Sun, Aug 16, 2020 at 1:36 AM Coly Li wrote:
>
> The original problem was from nvme-over-tcp code, who mistakenly uses
> kernel_sendpage() to send pages allocated by __get_free_pages() without
> __GFP_COMP flag. Such pages don't have refcount (page_count is 0) on
> tail pages, sending them by ke
The original problem was from nvme-over-tcp code, who mistakenly uses
kernel_sendpage() to send pages allocated by __get_free_pages() without
__GFP_COMP flag. Such pages don't have refcount (page_count is 0) on
tail pages, sending them by kernel_sendpage() may trigger a kernel panic
from a corrupte
The original problem was from nvme-over-tcp code, who mistakenly uses
kernel_sendpage() to send pages allocated by __get_free_pages() without
__GFP_COMP flag. Such pages don't have refcount (page_count is 0) on
tail pages, sending them by kernel_sendpage() may trigger a kernel panic
from a corrupte
