| 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
| ||
|
Message-ID: <4a1d7edcfae1e967eb2951f591c10c02965f6dc2.camel@redhat.com>
Date: Tue, 19 Sep 2023 10:09:53 +0200
From: Paolo Abeni <pabeni@...hat.com>
To: Eric Dumazet <edumazet@...gle.com>, Liang Chen
<liangchen.linux@...il.com>
Cc: davem@...emloft.net, kuba@...nel.org, benjamin.poirier@...il.com,
netdev@...r.kernel.org
Subject: Re: [PATCH net-next v4 2/2] pktgen: Introducing 'SHARED' flag for
testing with non-shared skb
On Mon, 2023-09-18 at 16:28 +0200, Eric Dumazet wrote:
> On Sat, Sep 16, 2023 at 3:30 PM Liang Chen <liangchen.linux@...il.com> wrote:
> >
> > Currently, skbs generated by pktgen always have their reference count
> > incremented before transmission, causing their reference count to be
> > always greater than 1, leading to two issues:
> > 1. Only the code paths for shared skbs can be tested.
> > 2. In certain situations, skbs can only be released by pktgen.
> > To enhance testing comprehensiveness, we are introducing the "SHARED"
> > flag to indicate whether an SKB is shared. This flag is enabled by
> > default, aligning with the current behavior. However, disabling this
> > flag allows skbs with a reference count of 1 to be transmitted.
> > So we can test non-shared skbs and code paths where skbs are released
> > within the stack.
> >
> > Signed-off-by: Liang Chen <liangchen.linux@...il.com>
> > ---
> > Documentation/networking/pktgen.rst | 12 ++++++++
> > net/core/pktgen.c | 48 ++++++++++++++++++++++++-----
> > 2 files changed, 52 insertions(+), 8 deletions(-)
> >
> > diff --git a/Documentation/networking/pktgen.rst b/Documentation/networking/pktgen.rst
> > index 1225f0f63ff0..c945218946e1 100644
> > --- a/Documentation/networking/pktgen.rst
> > +++ b/Documentation/networking/pktgen.rst
> > @@ -178,6 +178,7 @@ Examples::
> > IPSEC # IPsec encapsulation (needs CONFIG_XFRM)
> > NODE_ALLOC # node specific memory allocation
> > NO_TIMESTAMP # disable timestamping
> > + SHARED # enable shared SKB
> > pgset 'flag ![name]' Clear a flag to determine behaviour.
> > Note that you might need to use single quote in
> > interactive mode, so that your shell wouldn't expand
> > @@ -288,6 +289,16 @@ To avoid breaking existing testbed scripts for using AH type and tunnel mode,
> > you can use "pgset spi SPI_VALUE" to specify which transformation mode
> > to employ.
> >
> > +Disable shared SKB
> > +==================
> > +By default, SKBs sent by pktgen are shared (user count > 1).
> > +To test with non-shared SKBs, remove the "SHARED" flag by simply setting::
> > +
> > + pg_set "flag !SHARED"
> > +
> > +However, if the "clone_skb" or "burst" parameters are configured, the skb
> > +still needs to be held by pktgen for further access. Hence the skb must be
> > +shared.
> >
> > Current commands and configuration options
> > ==========================================
> > @@ -357,6 +368,7 @@ Current commands and configuration options
> > IPSEC
> > NODE_ALLOC
> > NO_TIMESTAMP
> > + SHARED
> >
> > spi (ipsec)
> >
> > diff --git a/net/core/pktgen.c b/net/core/pktgen.c
> > index 48306a101fd9..c4e0814df325 100644
> > --- a/net/core/pktgen.c
> > +++ b/net/core/pktgen.c
> > @@ -200,6 +200,7 @@
> > pf(VID_RND) /* Random VLAN ID */ \
> > pf(SVID_RND) /* Random SVLAN ID */ \
> > pf(NODE) /* Node memory alloc*/ \
> > + pf(SHARED) /* Shared SKB */ \
> >
> > #define pf(flag) flag##_SHIFT,
> > enum pkt_flags {
> > @@ -1198,7 +1199,8 @@ static ssize_t pktgen_if_write(struct file *file,
> > ((pkt_dev->xmit_mode == M_NETIF_RECEIVE) ||
> > !(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
> > return -ENOTSUPP;
> > - if (value > 0 && pkt_dev->n_imix_entries > 0)
> > + if (value > 0 && (pkt_dev->n_imix_entries > 0 ||
> > + !(pkt_dev->flags & F_SHARED)))
> > return -EINVAL;
> >
> > i += len;
> > @@ -1257,6 +1259,10 @@ static ssize_t pktgen_if_write(struct file *file,
> > ((pkt_dev->xmit_mode == M_START_XMIT) &&
> > (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))))
> > return -ENOTSUPP;
> > +
> > + if (value > 1 && !(pkt_dev->flags & F_SHARED))
> > + return -EINVAL;
> > +
> > pkt_dev->burst = value < 1 ? 1 : value;
> > sprintf(pg_result, "OK: burst=%u", pkt_dev->burst);
> > return count;
> > @@ -1334,10 +1340,19 @@ static ssize_t pktgen_if_write(struct file *file,
> >
> > flag = pktgen_read_flag(f, &disable);
> > if (flag) {
> > - if (disable)
> > + if (disable) {
> > + /* If "clone_skb", or "burst" parameters are
> > + * configured, it means that the skb still
> > + * needs to be referenced by the pktgen, so
> > + * the skb must be shared.
> > + */
> > + if (flag == F_SHARED && (pkt_dev->clone_skb ||
> > + pkt_dev->burst > 1))
> > + return -EINVAL;
> > pkt_dev->flags &= ~flag;
> > - else
> > + } else {
> > pkt_dev->flags |= flag;
> > + }
> >
> > sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
> > return count;
> > @@ -3489,7 +3504,8 @@ static void pktgen_xmit(struct pktgen_dev *pkt_dev)
> > if (pkt_dev->xmit_mode == M_NETIF_RECEIVE) {
> > skb = pkt_dev->skb;
> > skb->protocol = eth_type_trans(skb, skb->dev);
> > - refcount_add(burst, &skb->users);
> > + if (pkt_dev->flags & F_SHARED)
> > + refcount_add(burst, &skb->users);
> > local_bh_disable();
> > do {
> > ret = netif_receive_skb(skb);
> > @@ -3497,6 +3513,10 @@ static void pktgen_xmit(struct pktgen_dev *pkt_dev)
> > pkt_dev->errors++;
> > pkt_dev->sofar++;
> > pkt_dev->seq_num++;
>
> Since pkt_dev->flags can change under us, I would rather read pkt_dev->flags
> once in pktgen_xmit() to avoid surprises...
Additionally I *think* we can't assume pkt_dev->burst and pkt_dev-
>flags have consistent values in pktgen_xmit(). The user-space
(syzkaller) could flip burst and flag in between the read access in
pktgen_xmit().
There is a later:
if (unlikely(burst))
WARN_ON(refcount_sub_and_test(burst, &pkt_dev->skb->users));
that will need explicit check for 'pkt_dev->skb' not being NULL.
Cheers,
Paolo
Powered by blists - more mailing lists