Messages in this thread | | | Date | Wed, 8 May 2024 11:10:46 +0200 | Subject | Re: [PATCH net-next] netdevice: define and allocate &net_device _properly_ | From | Alexander Lobakin <> |
| |
From: Eric Dumazet <edumazet@google.com> Date: Tue, 7 May 2024 20:21:58 +0200
> On Tue, May 7, 2024 at 2:40 PM Alexander Lobakin > <aleksander.lobakin@intel.com> wrote: >> >> In fact, this structure contains a flexible array at the end, but >> historically its size, alignment etc., is calculated manually. >> There are several instances of the structure embedded into other >> structures, but also there's ongoing effort to remove them and we >> could in the meantime declare &net_device properly. >> Declare the array explicitly, use struct_size() and store the array >> size inside the structure, so that __counted_by() can be applied. >> Don't use PTR_ALIGN(), as SLUB itself tries its best to ensure the >> allocated buffer is aligned to what the user expects. >> Also, change its alignment from %NETDEV_ALIGN to the cacheline size >> as per several suggestions on the netdev ML. >> >> bloat-o-meter for vmlinux: >> >> free_netdev 445 440 -5 >> netdev_freemem 24 - -24 >> alloc_netdev_mqs 1481 1450 -31 >> >> On x86_64 with several NICs of different vendors, I was never able to >> get a &net_device pointer not aligned to the cacheline size after the >> change. >> >> Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com> >> --- > > ... > >> - p = kvzalloc(alloc_size, GFP_KERNEL_ACCOUNT | __GFP_RETRY_MAYFAIL); >> - if (!p) >> + sizeof_priv = ALIGN(sizeof_priv, SMP_CACHE_BYTES); > > If we have a __counted_by(priv_len), why do you ALIGN(sizeof_priv, > SMP_CACHE_BYTES) ?
To have the whole block of &net_device + private part aligned to a cacheline size.
> > If a driver pretends its private part is 4 bytes, we should get a > warning if 20 bytes are used instead.
Ah okay, so this should be
p = kvzalloc(struct_size(ALIGN(sizeof_priv, SMP_CACHE_BYTES))); p->priv_len = sizeof_priv;
> > You added two ____cacheline_aligned already in net_device already.
The whole size passed to kvzalloc() must be cacheline-aligned, otherwise the MM layer can miscalculate the pointer alignment.
Thanks, Olek
| |