linux-kernel - Re: [PATCH 3/4 RESEND] crypto: octeontx2: Fix address alignment on CN10K A0/A1 and OcteonTX2

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAAeCc_=QShbySa8x9zU+QnDqn1SLK3JLXMD8RYNoax+gh3NVEQ@mail.gmail.com>
Date: Mon, 19 May 2025 11:47:18 +0530
From: Bharat Bhushan <bharatb.linux@...il.com>
To: Herbert Xu <herbert@...dor.apana.org.au>
Cc: Bharat Bhushan <bbhushan2@...vell.com>, bbrezillon@...nel.org, arno@...isbad.org, 
	schalla@...vell.com, davem@...emloft.net, giovanni.cabiddu@...el.com, 
	linux@...blig.org, linux-crypto@...r.kernel.org, linux-kernel@...r.kernel.org, 
	stable@...r.kernel.org
Subject: Re: [PATCH 3/4 RESEND] crypto: octeontx2: Fix address alignment on
 CN10K A0/A1 and OcteonTX2

On Mon, May 19, 2025 at 9:57 AM Herbert Xu <herbert@...dor.apana.org.au> wrote:
>
> On Wed, May 14, 2025 at 10:40:42AM +0530, Bharat Bhushan wrote:
> >
> > @@ -429,22 +431,50 @@ otx2_sg_info_create(struct pci_dev *pdev, struct otx2_cpt_req_info *req,
> >               return NULL;
> >       }
> >
> > -     g_sz_bytes = ((req->in_cnt + 3) / 4) *
> > -                   sizeof(struct otx2_cpt_sglist_component);
> > -     s_sz_bytes = ((req->out_cnt + 3) / 4) *
> > -                   sizeof(struct otx2_cpt_sglist_component);
> > +     /* Allocate memory to meet below alignment requirement:
> > +      *  ----------------------------------
> > +      * |    struct otx2_cpt_inst_info     |
> > +      * |    (No alignment required)       |
> > +      * |     -----------------------------|
> > +      * |    | padding for 8B alignment    |
> > +      * |----------------------------------|
> > +      * |    SG List Gather/Input memory   |
> > +      * |    Length = multiple of 32Bytes  |
> > +      * |    Alignment = 8Byte             |
> > +      * |----------------------------------|
> > +      * |    SG List Scatter/Output memory |
> > +      * |    Length = multiple of 32Bytes  |
> > +      * |    Alignment = 8Byte             |
> > +      * |    (padding for below alignment) |
> > +      * |     -----------------------------|
> > +      * |    | padding for 32B alignment   |
> > +      * |----------------------------------|
> > +      * |    Result response memory        |
> > +      *  ----------------------------------
> > +      */
> >
> > -     dlen = g_sz_bytes + s_sz_bytes + SG_LIST_HDR_SIZE;
> > -     align_dlen = ALIGN(dlen, align);
> > -     info_len = ALIGN(sizeof(*info), align);
> > -     total_mem_len = align_dlen + info_len + sizeof(union otx2_cpt_res_s);
> > +     info_len = sizeof(*info);
> > +
> > +     g_len = ((req->in_cnt + 3) / 4) *
> > +              sizeof(struct otx2_cpt_sglist_component);
> > +     s_len = ((req->out_cnt + 3) / 4) *
> > +              sizeof(struct otx2_cpt_sglist_component);
> > +
> > +     dlen = g_len + s_len + SG_LIST_HDR_SIZE;
> > +
> > +     /* Allocate extra memory for SG and response address alignment */
> > +     total_mem_len = ALIGN(info_len, OTX2_CPT_DPTR_RPTR_ALIGN) + dlen;

This add extra memory for 8-byte (OTX2_CPT_DPTR_RPTR_ALIGN) alignment

> > +     total_mem_len = ALIGN(total_mem_len, OTX2_CPT_RES_ADDR_ALIGN) +
> > +                      sizeof(union otx2_cpt_res_s);

This add extra memory for 32-byte (OTX2_CPT_RES_ADDR_ALIGN))
In case not observed,  OTX2_CPT_RES_ADDR_ALIGN is not the same as
OTX2_CPT_DPTR_RPTR_ALIGN.

>
> This doesn't look right.  It would be correct if kzalloc returned
> a 32-byte aligned pointer to start with.  But it doesn't anymore,
> which is why you're making this patch in the first place :)
>
> So you need to add extra memory to bridge the gap between what it
> returns and what you expect.  Since it returns 8-byte aligned
> memory, and you expect 32-byte aligned pointers, you should add
> 24 bytes.
>
> IOW the calculation should be:
>
>         total_mem_len = ALIGN(info_len, OTX2_CPT_DPTR_RPTR_ALIGN) + dlen;
>         total_mem_len = ALIGN(total_mem_len, OTX2_CPT_DPTR_RPTR_ALIGN);
>         total_mem_len += (OTX2_CPT_RES_ADDR_ALIGN - 1) &
>                          ~(OTX2_CPT_DPTR_RPTR_ALIGN - 1);
>
> >       info = kzalloc(total_mem_len, gfp);
> >       if (unlikely(!info))
> >               return NULL;
> >
> >       info->dlen = dlen;
> > -     info->in_buffer = (u8 *)info + info_len;
> > +     info->in_buffer = PTR_ALIGN((u8 *)info + info_len,
> > +                                 OTX2_CPT_DPTR_RPTR_ALIGN);
> > +     info->out_buffer = info->in_buffer + 8 + g_len;
>
> I presume the 8 here corresponds to SG_LIST_HDR_SIZE from the dlen
> calculation above.  If so please spell it out as otherwise it's just
> confusing.

Yes, this is for SG_LIST_HDR_SIZE, will use same here.

Thanks
-Bharat

>
> Cheers,
> --
> Email: Herbert Xu <herbert@...dor.apana.org.au>
> Home Page: http://gondor.apana.org.au/~herbert/
> PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt