diff mbox

clnt_com_create: Restore backwards compatibility with the legacy glibc code.

Message ID 20180409172825.169380-1-steved@redhat.com (mailing list archive)
State New, archived
Headers show

Commit Message

Steve Dickson April 9, 2018, 5:28 p.m. UTC
Commit 46e04a73 changed clnt_com_create() to avoid
using reserved ports when creating the the CLIENT ptr.
This change breaks backward compatibility with the
legacy RPC code that was in glibc.

This patch reverts that commit to restore backwards compatibility

Signed-off-by: Steve Dickson <steved@redhat.com>
---
 src/rpc_soc.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

Comments

Chuck Lever April 9, 2018, 5:32 p.m. UTC | #1
> On Apr 9, 2018, at 11:28 AM, Steve Dickson <steved@redhat.com> wrote:
> 
> Commit 46e04a73 changed clnt_com_create() to avoid
> using reserved ports when creating the the CLIENT ptr.
> This change breaks backward compatibility with the
> legacy RPC code that was in glibc.
> 
> This patch reverts that commit to restore backwards compatibility
> 
> Signed-off-by: Steve Dickson <steved@redhat.com>
> ---
> src/rpc_soc.c | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/src/rpc_soc.c b/src/rpc_soc.c
> index af6c482..ed0892a 100644
> --- a/src/rpc_soc.c
> +++ b/src/rpc_soc.c
> @@ -67,8 +67,6 @@
> 
> extern mutex_t	rpcsoc_lock;
> 
> -extern int __binddynport(int fd);
> -
> static CLIENT *clnt_com_create(struct sockaddr_in *, rpcprog_t, rpcvers_t,
>     int *, u_int, u_int, char *, int);
> static SVCXPRT *svc_com_create(int, u_int, u_int, char *);
> @@ -147,8 +145,7 @@ clnt_com_create(raddr, prog, vers, sockp, sendsz, recvsz, tp, flags)
> 	bindaddr.maxlen = bindaddr.len =  sizeof (struct sockaddr_in);
> 	bindaddr.buf = raddr;
> 
> -	if (__binddynport(fd) == -1)
> -		goto err;
> +	bindresvport(fd, NULL);
> 	cl = clnt_tli_create(fd, nconf, &bindaddr, prog, vers,
> 				sendsz, recvsz);
> 	if (cl) {
> @@ -316,6 +313,7 @@ svc_com_create(fd, sendsize, recvsize, netid)
> 	SVCXPRT *svc;
> 	int madefd = FALSE;
> 	int port;
> +	struct sockaddr_in sin;
> 
> 	if ((nconf = __rpc_getconfip(netid)) == NULL) {
> 		(void) syslog(LOG_ERR, "Could not get %s transport", netid);
> @@ -332,6 +330,10 @@ svc_com_create(fd, sendsize, recvsize, netid)
> 		madefd = TRUE;
> 	}
> 
> +	memset(&sin, 0, sizeof sin);
> +	sin.sin_family = AF_INET;
> +	bindresvport(fd, &sin);
> +	listen(fd, SOMAXCONN);

Why do we need to fix the server API too?


> 	svc = svc_tli_create(fd, nconf, NULL, sendsize, recvsize);
> 	(void) freenetconfigent(nconf);
> 	if (svc == NULL) {
> -- 
> 2.14.3
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
Chuck Lever



--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Steve Dickson April 9, 2018, 5:58 p.m. UTC | #2
On 04/09/2018 01:32 PM, Chuck Lever wrote:
> 
> 
>> On Apr 9, 2018, at 11:28 AM, Steve Dickson <steved@redhat.com> wrote:
>>
>> Commit 46e04a73 changed clnt_com_create() to avoid
>> using reserved ports when creating the the CLIENT ptr.
>> This change breaks backward compatibility with the
>> legacy RPC code that was in glibc.
>>
>> This patch reverts that commit to restore backwards compatibility
>>
>> Signed-off-by: Steve Dickson <steved@redhat.com>
>> ---
>> src/rpc_soc.c | 10 ++++++----
>> 1 file changed, 6 insertions(+), 4 deletions(-)
>>
>> diff --git a/src/rpc_soc.c b/src/rpc_soc.c
>> index af6c482..ed0892a 100644
>> --- a/src/rpc_soc.c
>> +++ b/src/rpc_soc.c
>> @@ -67,8 +67,6 @@
>>
>> extern mutex_t	rpcsoc_lock;
>>
>> -extern int __binddynport(int fd);
>> -
>> static CLIENT *clnt_com_create(struct sockaddr_in *, rpcprog_t, rpcvers_t,
>>     int *, u_int, u_int, char *, int);
>> static SVCXPRT *svc_com_create(int, u_int, u_int, char *);
>> @@ -147,8 +145,7 @@ clnt_com_create(raddr, prog, vers, sockp, sendsz, recvsz, tp, flags)
>> 	bindaddr.maxlen = bindaddr.len =  sizeof (struct sockaddr_in);
>> 	bindaddr.buf = raddr;
>>
>> -	if (__binddynport(fd) == -1)
>> -		goto err;
>> +	bindresvport(fd, NULL);
>> 	cl = clnt_tli_create(fd, nconf, &bindaddr, prog, vers,
>> 				sendsz, recvsz);
>> 	if (cl) {
>> @@ -316,6 +313,7 @@ svc_com_create(fd, sendsize, recvsize, netid)
>> 	SVCXPRT *svc;
>> 	int madefd = FALSE;
>> 	int port;
>> +	struct sockaddr_in sin;
>>
>> 	if ((nconf = __rpc_getconfip(netid)) == NULL) {
>> 		(void) syslog(LOG_ERR, "Could not get %s transport", netid);
>> @@ -332,6 +330,10 @@ svc_com_create(fd, sendsize, recvsize, netid)
>> 		madefd = TRUE;
>> 	}
>>
>> +	memset(&sin, 0, sizeof sin);
>> +	sin.sin_family = AF_INET;
>> +	bindresvport(fd, &sin);
>> +	listen(fd, SOMAXCONN);
> 
> Why do we need to fix the server API too?
I thought about this... and I'm thinking there is
more of an exception of server listening on reserve
ports than clients using using reserve ports.

Plus this make us completely backwards compatibility
which in the long run I think is the right place to be.

steved.
 
> 
> 
>> 	svc = svc_tli_create(fd, nconf, NULL, sendsize, recvsize);
>> 	(void) freenetconfigent(nconf);
>> 	if (svc == NULL) {
>> -- 
>> 2.14.3
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> --
> Chuck Lever
> 
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Chuck Lever April 9, 2018, 6:40 p.m. UTC | #3
> On Apr 9, 2018, at 11:58 AM, Steve Dickson <SteveD@RedHat.com> wrote:
> 
> 
> 
> On 04/09/2018 01:32 PM, Chuck Lever wrote:
>> 
>> 
>>> On Apr 9, 2018, at 11:28 AM, Steve Dickson <steved@redhat.com> wrote:
>>> 
>>> Commit 46e04a73 changed clnt_com_create() to avoid
>>> using reserved ports when creating the the CLIENT ptr.
>>> This change breaks backward compatibility with the
>>> legacy RPC code that was in glibc.
>>> 
>>> This patch reverts that commit to restore backwards compatibility
>>> 
>>> Signed-off-by: Steve Dickson <steved@redhat.com>
>>> ---
>>> src/rpc_soc.c | 10 ++++++----
>>> 1 file changed, 6 insertions(+), 4 deletions(-)
>>> 
>>> diff --git a/src/rpc_soc.c b/src/rpc_soc.c
>>> index af6c482..ed0892a 100644
>>> --- a/src/rpc_soc.c
>>> +++ b/src/rpc_soc.c
>>> @@ -67,8 +67,6 @@
>>> 
>>> extern mutex_t	rpcsoc_lock;
>>> 
>>> -extern int __binddynport(int fd);
>>> -
>>> static CLIENT *clnt_com_create(struct sockaddr_in *, rpcprog_t, rpcvers_t,
>>>    int *, u_int, u_int, char *, int);
>>> static SVCXPRT *svc_com_create(int, u_int, u_int, char *);
>>> @@ -147,8 +145,7 @@ clnt_com_create(raddr, prog, vers, sockp, sendsz, recvsz, tp, flags)
>>> 	bindaddr.maxlen = bindaddr.len =  sizeof (struct sockaddr_in);
>>> 	bindaddr.buf = raddr;
>>> 
>>> -	if (__binddynport(fd) == -1)
>>> -		goto err;
>>> +	bindresvport(fd, NULL);
>>> 	cl = clnt_tli_create(fd, nconf, &bindaddr, prog, vers,
>>> 				sendsz, recvsz);
>>> 	if (cl) {
>>> @@ -316,6 +313,7 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>> 	SVCXPRT *svc;
>>> 	int madefd = FALSE;
>>> 	int port;
>>> +	struct sockaddr_in sin;
>>> 
>>> 	if ((nconf = __rpc_getconfip(netid)) == NULL) {
>>> 		(void) syslog(LOG_ERR, "Could not get %s transport", netid);
>>> @@ -332,6 +330,10 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>> 		madefd = TRUE;
>>> 	}
>>> 
>>> +	memset(&sin, 0, sizeof sin);
>>> +	sin.sin_family = AF_INET;
>>> +	bindresvport(fd, &sin);
>>> +	listen(fd, SOMAXCONN);
>> 
>> Why do we need to fix the server API too?
> I thought about this... and I'm thinking there is
> more of an exception of server listening on reserve
> ports than clients using using reserve ports.

Hi Steve-

I think you mean that there are fewer legacy servers
than there are legacy clients? I don't see how that
matters: There's no benefit at all for having a
server listen on a dynamically-assigned privileged
port. Nothing can be broken if the server API uses
an ephemeral port.


> Plus this make us completely backwards compatibility
> which in the long run I think is the right place to be.

Philosophically I agree that backwards compatibility
is good, but I think you're "going to hell with the
joke" as they say.

The point of backwards compatibility is that we don't
want to break applications that depend on some behavior.
There can't be any applications that depend on this
behavior because there's no functional benefit to it,
only a downside, which we have no reason not to fix.

I'm going to firmly object on this one, unless you have
a clearly documented breakage that requires the legacy
server API to use bindresvport(3).

Also, it would be great to get a man page update on the
legacy clnt API that documents the behavior when a root
caller uses that API. That gives some guarantee that it
doesn't get changed again inadvertently.


> steved.
> 
>> 
>> 
>>> 	svc = svc_tli_create(fd, nconf, NULL, sendsize, recvsize);
>>> 	(void) freenetconfigent(nconf);
>>> 	if (svc == NULL) {
>>> -- 
>>> 2.14.3
>>> 
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> 
>> --
>> Chuck Lever

--
Chuck Lever



--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Steve Dickson April 9, 2018, 8:04 p.m. UTC | #4
On 04/09/2018 02:40 PM, Chuck Lever wrote:
> 
> 
>> On Apr 9, 2018, at 11:58 AM, Steve Dickson <SteveD@RedHat.com> wrote:
>>
>>
>>
>> On 04/09/2018 01:32 PM, Chuck Lever wrote:
>>>
>>>
>>>> On Apr 9, 2018, at 11:28 AM, Steve Dickson <steved@redhat.com> wrote:
>>>>
>>>> Commit 46e04a73 changed clnt_com_create() to avoid
>>>> using reserved ports when creating the the CLIENT ptr.
>>>> This change breaks backward compatibility with the
>>>> legacy RPC code that was in glibc.
>>>>
>>>> This patch reverts that commit to restore backwards compatibility
>>>>
>>>> Signed-off-by: Steve Dickson <steved@redhat.com>
>>>> ---
>>>> src/rpc_soc.c | 10 ++++++----
>>>> 1 file changed, 6 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/src/rpc_soc.c b/src/rpc_soc.c
>>>> index af6c482..ed0892a 100644
>>>> --- a/src/rpc_soc.c
>>>> +++ b/src/rpc_soc.c
>>>> @@ -67,8 +67,6 @@
>>>>
>>>> extern mutex_t	rpcsoc_lock;
>>>>
>>>> -extern int __binddynport(int fd);
>>>> -
>>>> static CLIENT *clnt_com_create(struct sockaddr_in *, rpcprog_t, rpcvers_t,
>>>>    int *, u_int, u_int, char *, int);
>>>> static SVCXPRT *svc_com_create(int, u_int, u_int, char *);
>>>> @@ -147,8 +145,7 @@ clnt_com_create(raddr, prog, vers, sockp, sendsz, recvsz, tp, flags)
>>>> 	bindaddr.maxlen = bindaddr.len =  sizeof (struct sockaddr_in);
>>>> 	bindaddr.buf = raddr;
>>>>
>>>> -	if (__binddynport(fd) == -1)
>>>> -		goto err;
>>>> +	bindresvport(fd, NULL);
>>>> 	cl = clnt_tli_create(fd, nconf, &bindaddr, prog, vers,
>>>> 				sendsz, recvsz);
>>>> 	if (cl) {
>>>> @@ -316,6 +313,7 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>>> 	SVCXPRT *svc;
>>>> 	int madefd = FALSE;
>>>> 	int port;
>>>> +	struct sockaddr_in sin;
>>>>
>>>> 	if ((nconf = __rpc_getconfip(netid)) == NULL) {
>>>> 		(void) syslog(LOG_ERR, "Could not get %s transport", netid);
>>>> @@ -332,6 +330,10 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>>> 		madefd = TRUE;
>>>> 	}
>>>>
>>>> +	memset(&sin, 0, sizeof sin);
>>>> +	sin.sin_family = AF_INET;
>>>> +	bindresvport(fd, &sin);
>>>> +	listen(fd, SOMAXCONN);
>>>
>>> Why do we need to fix the server API too?
>> I thought about this... and I'm thinking there is
>> more of an exception of server listening on reserve
>> ports than clients using using reserve ports.
> 
> Hi Steve-
> 
> I think you mean that there are fewer legacy servers
> than there are legacy clients? I don't see how that
> matters: There's no benefit at all for having a
> server listen on a dynamically-assigned privileged
> port. Nothing can be broken if the server API uses
> an ephemeral port.
> 
> 
>> Plus this make us completely backwards compatibility
>> which in the long run I think is the right place to be.
> 
> Philosophically I agree that backwards compatibility
> is good, but I think you're "going to hell with the
> joke" as they say.
> 
> The point of backwards compatibility is that we don't
> want to break applications that depend on some behavior.
> There can't be any applications that depend on this
> behavior because there's no functional benefit to it,
> only a downside, which we have no reason not to fix.The downside is we are changing an API that has 
been established for since the 80's... What good
could that possibly do WRT legacy servers?
 
> 
> I'm going to firmly object on this one, unless you have
> a clearly documented breakage that requires the legacy
> server API to use bindresvport(3).
Here is what the man page says "If the socket is not bound to a 
local TCP port, then this routine binds it to an arbitrary port."

The point being it also does not talk about creating a 
listening connection either... Changing the (undocumented)
API like this can cause nothing but problems... IMHO...

Basically, not making this change will cause a fork with 
all the major distros since it very import for them to be 
backward compatible esp in enterprise worlds. Upstream not 
so much... Who really uses raw upstream bits in a stable
environment... understood... But this brings me to my point.

What problem is being fixed by changing an 20+ year API? 
Where are the bug reports that this change is needed
or wanted? 

> 
> Also, it would be great to get a man page update on the
> legacy clnt API that documents the behavior when a root
> caller uses that API. That gives some guarantee that it
> doesn't get changed again inadvertently.
I need to look, but I think glibc still "owns" the
man pages... 

steved.
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Chuck Lever April 9, 2018, 9:11 p.m. UTC | #5
> On Apr 9, 2018, at 2:04 PM, Steve Dickson <SteveD@RedHat.com> wrote:
> 
> 
> 
> On 04/09/2018 02:40 PM, Chuck Lever wrote:
>> 
>> 
>>> On Apr 9, 2018, at 11:58 AM, Steve Dickson <SteveD@RedHat.com> wrote:
>>> 
>>> 
>>> 
>>> On 04/09/2018 01:32 PM, Chuck Lever wrote:
>>>> 
>>>> 
>>>>> On Apr 9, 2018, at 11:28 AM, Steve Dickson <steved@redhat.com> wrote:
>>>>> 
>>>>> Commit 46e04a73 changed clnt_com_create() to avoid
>>>>> using reserved ports when creating the the CLIENT ptr.
>>>>> This change breaks backward compatibility with the
>>>>> legacy RPC code that was in glibc.
>>>>> 
>>>>> This patch reverts that commit to restore backwards compatibility
>>>>> 
>>>>> Signed-off-by: Steve Dickson <steved@redhat.com>
>>>>> ---
>>>>> src/rpc_soc.c | 10 ++++++----
>>>>> 1 file changed, 6 insertions(+), 4 deletions(-)
>>>>> 
>>>>> diff --git a/src/rpc_soc.c b/src/rpc_soc.c
>>>>> index af6c482..ed0892a 100644
>>>>> --- a/src/rpc_soc.c
>>>>> +++ b/src/rpc_soc.c
>>>>> @@ -67,8 +67,6 @@
>>>>> 
>>>>> extern mutex_t	rpcsoc_lock;
>>>>> 
>>>>> -extern int __binddynport(int fd);
>>>>> -
>>>>> static CLIENT *clnt_com_create(struct sockaddr_in *, rpcprog_t, rpcvers_t,
>>>>>   int *, u_int, u_int, char *, int);
>>>>> static SVCXPRT *svc_com_create(int, u_int, u_int, char *);
>>>>> @@ -147,8 +145,7 @@ clnt_com_create(raddr, prog, vers, sockp, sendsz, recvsz, tp, flags)
>>>>> 	bindaddr.maxlen = bindaddr.len =  sizeof (struct sockaddr_in);
>>>>> 	bindaddr.buf = raddr;
>>>>> 
>>>>> -	if (__binddynport(fd) == -1)
>>>>> -		goto err;
>>>>> +	bindresvport(fd, NULL);
>>>>> 	cl = clnt_tli_create(fd, nconf, &bindaddr, prog, vers,
>>>>> 				sendsz, recvsz);
>>>>> 	if (cl) {
>>>>> @@ -316,6 +313,7 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>>>> 	SVCXPRT *svc;
>>>>> 	int madefd = FALSE;
>>>>> 	int port;
>>>>> +	struct sockaddr_in sin;
>>>>> 
>>>>> 	if ((nconf = __rpc_getconfip(netid)) == NULL) {
>>>>> 		(void) syslog(LOG_ERR, "Could not get %s transport", netid);
>>>>> @@ -332,6 +330,10 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>>>> 		madefd = TRUE;
>>>>> 	}
>>>>> 
>>>>> +	memset(&sin, 0, sizeof sin);
>>>>> +	sin.sin_family = AF_INET;
>>>>> +	bindresvport(fd, &sin);
>>>>> +	listen(fd, SOMAXCONN);
>>>> 
>>>> Why do we need to fix the server API too?
>>> I thought about this... and I'm thinking there is
>>> more of an exception of server listening on reserve
>>> ports than clients using using reserve ports.
>> 
>> Hi Steve-
>> 
>> I think you mean that there are fewer legacy servers
>> than there are legacy clients? I don't see how that
>> matters: There's no benefit at all for having a
>> server listen on a dynamically-assigned privileged
>> port. Nothing can be broken if the server API uses
>> an ephemeral port.
>> 
>> 
>>> Plus this make us completely backwards compatibility
>>> which in the long run I think is the right place to be.
>> 
>> Philosophically I agree that backwards compatibility
>> is good, but I think you're "going to hell with the
>> joke" as they say.
>> 
>> The point of backwards compatibility is that we don't
>> want to break applications that depend on some behavior.
>> There can't be any applications that depend on this
>> behavior because there's no functional benefit to it,
>> only a downside, which we have no reason not to fix.
> The downside is we are changing an API that has 
> been established for since the 80's... What good
> could that possibly do WRT legacy servers?

That's the wrong question, IMO. It won't hurt legacy
servers at all, and will benefit everyone else.


>> I'm going to firmly object on this one, unless you have
>> a clearly documented breakage that requires the legacy
>> server API to use bindresvport(3).
> Here is what the man page says "If the socket is not bound to a 
> local TCP port, then this routine binds it to an arbitrary port."
> 
> The point being it also does not talk about creating a 
> listening connection either... Changing the (undocumented)
> API like this can cause nothing but problems... IMHO...

Please show me one specific breakage that will result
in removing bindresvport from svc_com_create. Just one,
and I promise I will crawl back into my hole.

In case it wasn't clear, my preference is for a patch
that reverts only the removal of bindresvport(3) from
clnt_com_create. The svc_com_create change is safe and
should remain unreverted.

I don't like reverting the clnt_com_create change, but
I won't object to it, and there is clear evidence that
some old programs are inconvenienced by that change.


> Basically, not making this change will cause a fork with 
> all the major distros since it very import for them to be 
> backward compatible esp in enterprise worlds.

That sounds like an overstatement. Who claims they won't
take libtirpc with an unprivileged svc_com_create? I would
really like to understand why.


> Upstream not 
> so much... Who really uses raw upstream bits in a stable
> environment... understood... But this brings me to my point.
> 
> What problem is being fixed by changing an 20+ year API? 

The problem is legacy clients and servers are squatting on
ports that are assigned to other network services. These
patches mitigate that problem. There is more to be done.

The backwards compatibility issue is that some old servers
believe that clients have to use a privileged source port
to show that they are an authorized RPC consumer. We address
that by reverting clnt_com_create to use bindresvport(3),
although we can easily decide this is a security bug that
is worth forcing legacy API consumers to fix their usage.

There is no similar backwards compatibility issue for
clients talking to servers. That's just a fact about how
RPC operates: the client uses rpcbind to discover the
server's port. After that, there's no additional check to
see if that service port is less than 1024.


> Where are the bug reports that this change is needed
> or wanted? 

I was presented with a list of five or more bug reports
from various distributors that go back to almost 2000
where RPC consumers have caused issues occupying IANA-
assigned reserved ports.

https://sourceforge.net/p/libtirpc/mailman/libtirpc-devel/?viewmonth=201801&viewday=12&style=flat

The bottom item on that page is Guillem's e-mail that
cites bug reports on this issue.


>> Also, it would be great to get a man page update on the
>> legacy clnt API that documents the behavior when a root
>> caller uses that API. That gives some guarantee that it
>> doesn't get changed again inadvertently.
> I need to look, but I think glibc still "owns" the
> man pages... 

That should be fixed if that's true. Since RPC is to be
removed from glibc I can't think why they'd want to keep
the RPC man pages.

However, libtirpc has a bunch of man pages for the legacy
APIs. At least those can be updated.


--
Chuck Lever



--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Steve Dickson April 10, 2018, 3:21 p.m. UTC | #6
On 04/09/2018 05:11 PM, Chuck Lever wrote:
> 
> 
>> On Apr 9, 2018, at 2:04 PM, Steve Dickson <SteveD@RedHat.com> wrote:
>>
>>
>>
>> On 04/09/2018 02:40 PM, Chuck Lever wrote:
>>>
>>>
>>>> On Apr 9, 2018, at 11:58 AM, Steve Dickson <SteveD@RedHat.com> wrote:
>>>>
>>>>
>>>>
>>>> On 04/09/2018 01:32 PM, Chuck Lever wrote:
>>>>>
>>>>>
>>>>>> On Apr 9, 2018, at 11:28 AM, Steve Dickson <steved@redhat.com> wrote:
>>>>>>
>>>>>> Commit 46e04a73 changed clnt_com_create() to avoid
>>>>>> using reserved ports when creating the the CLIENT ptr.
>>>>>> This change breaks backward compatibility with the
>>>>>> legacy RPC code that was in glibc.
>>>>>>
>>>>>> This patch reverts that commit to restore backwards compatibility
>>>>>>
>>>>>> Signed-off-by: Steve Dickson <steved@redhat.com>
>>>>>> ---
>>>>>> src/rpc_soc.c | 10 ++++++----
>>>>>> 1 file changed, 6 insertions(+), 4 deletions(-)
>>>>>>
>>>>>> diff --git a/src/rpc_soc.c b/src/rpc_soc.c
>>>>>> index af6c482..ed0892a 100644
>>>>>> --- a/src/rpc_soc.c
>>>>>> +++ b/src/rpc_soc.c
>>>>>> @@ -67,8 +67,6 @@
>>>>>>
>>>>>> extern mutex_t	rpcsoc_lock;
>>>>>>
>>>>>> -extern int __binddynport(int fd);
>>>>>> -
>>>>>> static CLIENT *clnt_com_create(struct sockaddr_in *, rpcprog_t, rpcvers_t,
>>>>>>   int *, u_int, u_int, char *, int);
>>>>>> static SVCXPRT *svc_com_create(int, u_int, u_int, char *);
>>>>>> @@ -147,8 +145,7 @@ clnt_com_create(raddr, prog, vers, sockp, sendsz, recvsz, tp, flags)
>>>>>> 	bindaddr.maxlen = bindaddr.len =  sizeof (struct sockaddr_in);
>>>>>> 	bindaddr.buf = raddr;
>>>>>>
>>>>>> -	if (__binddynport(fd) == -1)
>>>>>> -		goto err;
>>>>>> +	bindresvport(fd, NULL);
>>>>>> 	cl = clnt_tli_create(fd, nconf, &bindaddr, prog, vers,
>>>>>> 				sendsz, recvsz);
>>>>>> 	if (cl) {
>>>>>> @@ -316,6 +313,7 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>>>>> 	SVCXPRT *svc;
>>>>>> 	int madefd = FALSE;
>>>>>> 	int port;
>>>>>> +	struct sockaddr_in sin;
>>>>>>
>>>>>> 	if ((nconf = __rpc_getconfip(netid)) == NULL) {
>>>>>> 		(void) syslog(LOG_ERR, "Could not get %s transport", netid);
>>>>>> @@ -332,6 +330,10 @@ svc_com_create(fd, sendsize, recvsize, netid)
>>>>>> 		madefd = TRUE;
>>>>>> 	}
>>>>>>
>>>>>> +	memset(&sin, 0, sizeof sin);
>>>>>> +	sin.sin_family = AF_INET;
>>>>>> +	bindresvport(fd, &sin);
>>>>>> +	listen(fd, SOMAXCONN);
>>>>>
>>>>> Why do we need to fix the server API too?
>>>> I thought about this... and I'm thinking there is
>>>> more of an exception of server listening on reserve
>>>> ports than clients using using reserve ports.
>>>
>>> Hi Steve-
>>>
>>> I think you mean that there are fewer legacy servers
>>> than there are legacy clients? I don't see how that
>>> matters: There's no benefit at all for having a
>>> server listen on a dynamically-assigned privileged
>>> port. Nothing can be broken if the server API uses
>>> an ephemeral port.
>>>
>>>
>>>> Plus this make us completely backwards compatibility
>>>> which in the long run I think is the right place to be.
>>>
>>> Philosophically I agree that backwards compatibility
>>> is good, but I think you're "going to hell with the
>>> joke" as they say.
>>>
>>> The point of backwards compatibility is that we don't
>>> want to break applications that depend on some behavior.
>>> There can't be any applications that depend on this
>>> behavior because there's no functional benefit to it,
>>> only a downside, which we have no reason not to fix.
>> The downside is we are changing an API that has 
>> been established for since the 80's... What good
>> could that possibly do WRT legacy servers?
> 
> That's the wrong question, IMO. It won't hurt legacy
> servers at all, and will benefit everyone else.
> 
> 
>>> I'm going to firmly object on this one, unless you have
>>> a clearly documented breakage that requires the legacy
>>> server API to use bindresvport(3).
>> Here is what the man page says "If the socket is not bound to a 
>> local TCP port, then this routine binds it to an arbitrary port."
>>
>> The point being it also does not talk about creating a 
>> listening connection either... Changing the (undocumented)
>> API like this can cause nothing but problems... IMHO...
> 
> Please show me one specific breakage that will result
> in removing bindresvport from svc_com_create. Just one,
> and I promise I will crawl back into my hole.
The patch also restores the listen() which is
something I can see  a server apps depending on. 

> 
> In case it wasn't clear, my preference is for a patch
> that reverts only the removal of bindresvport(3) from
> clnt_com_create. The svc_com_create change is safe and
> should remain unreverted.
> 
> I don't like reverting the clnt_com_create change, but
> I won't object to it, and there is clear evidence that
> some old programs are inconvenienced by that change.
I do not want to be in that position again... 
I just don't want to wait around to see what breaks.

> 
> 
>> Basically, not making this change will cause a fork with 
>> all the major distros since it very import for them to be 
>> backward compatible esp in enterprise worlds.
> 
> That sounds like an overstatement. Who claims they won't
> take libtirpc with an unprivileged svc_com_create? I would
> really like to understand why.
I'm just assuming most distros want to be backward compatible
and not break legacy apps... Maybe I'm wrong... 
 
> 
> 
>> Upstream not 
>> so much... Who really uses raw upstream bits in a stable
>> environment... understood... But this brings me to my point.
>>
>> What problem is being fixed by changing an 20+ year API? 
> 
> The problem is legacy clients and servers are squatting on
> ports that are assigned to other network services. These
> patches mitigate that problem. There is more to be done.
Any examples of these... The only one I know is rpcbind
which can be fixed in another ways.... 

> 
> The backwards compatibility issue is that some old servers
> believe that clients have to use a privileged source port
> to show that they are an authorized RPC consumer. We address
> that by reverting clnt_com_create to use bindresvport(3),
> although we can easily decide this is a security bug that
> is worth forcing legacy API consumers to fix their usage.
> 
> There is no similar backwards compatibility issue for
> clients talking to servers. That's just a fact about how
> RPC operates: the client uses rpcbind to discover the
> server's port. After that, there's no additional check to
> see if that service port is less than 1024.
Well its going to be hard for the client to talk
to the server if the server is no longer listening ;-)

> 
> 
>> Where are the bug reports that this change is needed
>> or wanted? 
> 
> I was presented with a list of five or more bug reports
> from various distributors that go back to almost 2000
> where RPC consumers have caused issues occupying IANA-
> assigned reserved ports.
> 
> https://sourceforge.net/p/libtirpc/mailman/libtirpc-devel/?viewmonth=201801&viewday=12&style=flat
> 
> The bottom item on that page is Guillem's e-mail that
> cites bug reports on this issue.
> 
> 
>>> Also, it would be great to get a man page update on the
>>> legacy clnt API that documents the behavior when a root
>>> caller uses that API. That gives some guarantee that it
>>> doesn't get changed again inadvertently.
>> I need to look, but I think glibc still "owns" the
>> man pages... 
> 
> That should be fixed if that's true. Since RPC is to be
> removed from glibc I can't think why they'd want to keep
> the RPC man pages.
I'll talk to the glibc people to what they are planning... 

steved.
> 
> However, libtirpc has a bunch of man pages for the legacy
> APIs. At least those can be updated.
> 
> 
> --
> Chuck Lever
> 
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Chuck Lever April 10, 2018, 10:43 p.m. UTC | #7
> On Apr 10, 2018, at 9:21 AM, Steve Dickson <SteveD@RedHat.com> wrote:
> 
> On 04/09/2018 05:11 PM, Chuck Lever wrote:
>> 
>>> On Apr 9, 2018, at 2:04 PM, Steve Dickson <SteveD@RedHat.com> wrote:
>>> 
>>> On 04/09/2018 02:40 PM, Chuck Lever wrote:
>>>> I'm going to firmly object on this one, unless you have
>>>> a clearly documented breakage that requires the legacy
>>>> server API to use bindresvport(3).
>>> Here is what the man page says "If the socket is not bound to a 
>>> local TCP port, then this routine binds it to an arbitrary port."
>>> 
>>> The point being it also does not talk about creating a 
>>> listening connection either... Changing the (undocumented)
>>> API like this can cause nothing but problems... IMHO...
>> 
>> Please show me one specific breakage that will result
>> in removing bindresvport from svc_com_create. Just one,
>> and I promise I will crawl back into my hole.
> The patch also restores the listen() which is
> something I can see  a server apps depending on. 

I'm not clear there's a problem there. Did you find
that svc_tli_create() was not already doing a listen?


>>> Basically, not making this change will cause a fork with 
>>> all the major distros since it very import for them to be 
>>> backward compatible esp in enterprise worlds.
>> 
>> That sounds like an overstatement. Who claims they won't
>> take libtirpc with an unprivileged svc_com_create? I would
>> really like to understand why.
> I'm just assuming most distros want to be backward compatible
> and not break legacy apps... Maybe I'm wrong...

Of course they don't want to break legacy apps. I just don't
see any backwards compatibility requirement here: servers
never need a reserved port, they need either a specific port,
or any port.

Therefore the libtirpc server-side APIs should not use
bindresvport.


>>> Upstream not 
>>> so much... Who really uses raw upstream bits in a stable
>>> environment... understood... But this brings me to my point.
>>> 
>>> What problem is being fixed by changing an 20+ year API? 
>> 
>> The problem is legacy clients and servers are squatting on
>> ports that are assigned to other network services. These
>> patches mitigate that problem. There is more to be done.
> Any examples of these... The only one I know is rpcbind
> which can be fixed in another ways.... 

Yes, as you and I discussed with Trond, statd can squat
on a reserved port for it's downcall client to lockd. The
fix Trond prefers is for this downcall service to use an
AF_LOCAL socket instead.

We should also update bindresvport(3) to do more to avoid
selecting ports that appear in /etc/services. This will
cover user apps that use the RPC library API.


--
Chuck Lever



--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Steve Dickson April 11, 2018, 2:10 p.m. UTC | #8
On 04/10/2018 06:43 PM, Chuck Lever wrote:
> 
> 
>> On Apr 10, 2018, at 9:21 AM, Steve Dickson <SteveD@RedHat.com> wrote:
>>
>> On 04/09/2018 05:11 PM, Chuck Lever wrote:
>>>
>>>> On Apr 9, 2018, at 2:04 PM, Steve Dickson <SteveD@RedHat.com> wrote:
>>>>
>>>> On 04/09/2018 02:40 PM, Chuck Lever wrote:
>>>>> I'm going to firmly object on this one, unless you have
>>>>> a clearly documented breakage that requires the legacy
>>>>> server API to use bindresvport(3).
>>>> Here is what the man page says "If the socket is not bound to a 
>>>> local TCP port, then this routine binds it to an arbitrary port."
>>>>
>>>> The point being it also does not talk about creating a 
>>>> listening connection either... Changing the (undocumented)
>>>> API like this can cause nothing but problems... IMHO...
>>>
>>> Please show me one specific breakage that will result
>>> in removing bindresvport from svc_com_create. Just one,
>>> and I promise I will crawl back into my hole.
>> The patch also restores the listen() which is
>> something I can see  a server apps depending on. 
> 
> I'm not clear there's a problem there. Did you find
> that svc_tli_create() was not already doing a listen?
yeah I didn't see that... 

> 
> 
>>>> Basically, not making this change will cause a fork with 
>>>> all the major distros since it very import for them to be 
>>>> backward compatible esp in enterprise worlds.
>>>
>>> That sounds like an overstatement. Who claims they won't
>>> take libtirpc with an unprivileged svc_com_create? I would
>>> really like to understand why.
>> I'm just assuming most distros want to be backward compatible
>> and not break legacy apps... Maybe I'm wrong...
> 
> Of course they don't want to break legacy apps. I just don't
> see any backwards compatibility requirement here: servers
> never need a reserved port, they need either a specific port,
> or any port.
> 
> Therefore the libtirpc server-side APIs should not use
> bindresvport.
> 
> 
>>>> Upstream not 
>>>> so much... Who really uses raw upstream bits in a stable
>>>> environment... understood... But this brings me to my point.
>>>>
>>>> What problem is being fixed by changing an 20+ year API? 
>>>
>>> The problem is legacy clients and servers are squatting on
>>> ports that are assigned to other network services. These
>>> patches mitigate that problem. There is more to be done.
>> Any examples of these... The only one I know is rpcbind
>> which can be fixed in another ways.... 
> 
> Yes, as you and I discussed with Trond, statd can squat
> on a reserved port for it's downcall client to lockd. The
> fix Trond prefers is for this downcall service to use an
> AF_LOCAL socket instead.
But it is careful not to used any ports in /etc/services

steved.

> 
> We should also update bindresvport(3) to do more to avoid
> selecting ports that appear in /etc/services. This will
> cover user apps that use the RPC library API.
> 
> 
> --
> Chuck Lever
> 
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Chuck Lever April 11, 2018, 3:09 p.m. UTC | #9
> On Apr 11, 2018, at 8:10 AM, Steve Dickson <SteveD@RedHat.com> wrote:
> 
> 
> 
> On 04/10/2018 06:43 PM, Chuck Lever wrote:
>> 
>> 
>>> On Apr 10, 2018, at 9:21 AM, Steve Dickson <SteveD@RedHat.com> wrote:
>>> 
>>>>> Upstream not 
>>>>> so much... Who really uses raw upstream bits in a stable
>>>>> environment... understood... But this brings me to my point.
>>>>> 
>>>>> What problem is being fixed by changing an 20+ year API? 
>>>> 
>>>> The problem is legacy clients and servers are squatting on
>>>> ports that are assigned to other network services. These
>>>> patches mitigate that problem. There is more to be done.
>>> Any examples of these... The only one I know is rpcbind
>>> which can be fixed in another ways.... 
>> 
>> Yes, as you and I discussed with Trond, statd can squat
>> on a reserved port for it's downcall client to lockd. The
>> fix Trond prefers is for this downcall service to use an
>> AF_LOCAL socket instead.
> But it is careful not to used any ports in /etc/services

There are two sides to this reserved port coin.

- The "heads" side is that long-lived clients shouldn't squat
on a port that is allocated to another service. This includes
IANA-allocated ports below and above 1024.

- The "tails" side is that avoiding IANA-allocated ports below
1024 reduces the already-small pool of reserved ports that can
be selected.

The "heads" part can be addressed by making bindresvport(3)
act more like the current statd downcall client. Even if many
clients are short-lived, we don't have any control over how
long clients live, and even a short-lived client can block
reliable start-up of well-known services temporarily, leading
to annoying non-determinant behavior. Thus making this API
work better is still important. (This is what Guillem was
trying to address in January).

Moreover we have to ensure that bindresvport(3) doesn't fail
to find a port just because the only ports that happen to be
available are IANA-allocated. Thus we should address the
"tails" part of this too, and make libtirpc uses this scarce
shared resource economically. (This is the part Guillem's
patch didn't address, as many reviewers pointed out).

There are some applications that use reserved ports today that
don't need to:

 - Any listener, for example, does not need an arbitrarily-
   assigned reserved port. By using __binddynport, the
   dynamically assigned port avoids all IANA-allocated ports
   (even IANA-allocated ports above 1024).

 - Many client applications, even if they are short-lived, do
   not require a privileged source port. And to avoid an IANA-
   allocated port, they should likewise use __binddynport.
   That really should be the default behavior of the CLNT
   interfaces, because bindresvport(3) is available for the
   few clients that need a privileged source port.

 - As you pointed out, applications that use TCP will leave a
   port in TIME_WAIT, which takes that source port out of the
   pool for 120 seconds. That's painful if the port is reserved.

Fixing these applications helps to address "tails". By changing
the libtirpc client APIs, we address the second and third points
for a fleet of RPC clients.

I don't have much empathy for YP:
- The use of NIS/YP is deprecated
- The use of a privileged source port is security theater
- The API behavior YP relies on is not documented
- There already exist patches available to make YP do this right

In this particular case, although backwards compatibility is
important, IMO reliable system operation is a higher priority.

--
Chuck Lever



--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/src/rpc_soc.c b/src/rpc_soc.c
index af6c482..ed0892a 100644
--- a/src/rpc_soc.c
+++ b/src/rpc_soc.c
@@ -67,8 +67,6 @@ 
 
 extern mutex_t	rpcsoc_lock;
 
-extern int __binddynport(int fd);
-
 static CLIENT *clnt_com_create(struct sockaddr_in *, rpcprog_t, rpcvers_t,
     int *, u_int, u_int, char *, int);
 static SVCXPRT *svc_com_create(int, u_int, u_int, char *);
@@ -147,8 +145,7 @@  clnt_com_create(raddr, prog, vers, sockp, sendsz, recvsz, tp, flags)
 	bindaddr.maxlen = bindaddr.len =  sizeof (struct sockaddr_in);
 	bindaddr.buf = raddr;
 
-	if (__binddynport(fd) == -1)
-		goto err;
+	bindresvport(fd, NULL);
 	cl = clnt_tli_create(fd, nconf, &bindaddr, prog, vers,
 				sendsz, recvsz);
 	if (cl) {
@@ -316,6 +313,7 @@  svc_com_create(fd, sendsize, recvsize, netid)
 	SVCXPRT *svc;
 	int madefd = FALSE;
 	int port;
+	struct sockaddr_in sin;
 
 	if ((nconf = __rpc_getconfip(netid)) == NULL) {
 		(void) syslog(LOG_ERR, "Could not get %s transport", netid);
@@ -332,6 +330,10 @@  svc_com_create(fd, sendsize, recvsize, netid)
 		madefd = TRUE;
 	}
 
+	memset(&sin, 0, sizeof sin);
+	sin.sin_family = AF_INET;
+	bindresvport(fd, &sin);
+	listen(fd, SOMAXCONN);
 	svc = svc_tli_create(fd, nconf, NULL, sendsize, recvsize);
 	(void) freenetconfigent(nconf);
 	if (svc == NULL) {