This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] powerpc: New feature - HWCAP/HWCAP2 bits in the TCB
- From: Rich Felker <dalias at libc dot org>
- To: munroesj at linux dot vnet dot ibm dot com
- Cc: Carlos O'Donell <carlos at redhat dot com>, libc-alpha at sourceware dot org
- Date: Mon, 6 Jul 2015 17:56:57 -0400
- Subject: Re: [PATCH] powerpc: New feature - HWCAP/HWCAP2 bits in the TCB
- Authentication-results: sourceware.org; auth=none
- References: <55760314 dot 6070601 at linux dot vnet dot ibm dot com> <20150609163835 dot GI17573 at brightrain dot aerifal dot cx> <1435777940 dot 7125 dot 132 dot camel at oc7878010663> <5596C284 dot 9070108 at redhat dot com> <1436145404 dot 10792 dot 46 dot camel at oc7878010663> <20150706021300 dot GU1173 at brightrain dot aerifal dot cx> <1436189206 dot 9162 dot 21 dot camel at oc7878010663> <20150706155219 dot GV1173 at brightrain dot aerifal dot cx> <1436217981 dot 8449 dot 8 dot camel at oc7878010663>
On Mon, Jul 06, 2015 at 04:26:21PM -0500, Steven Munroe wrote:
> > > The dword we are talking about is already allocated and has been since
> > > the initial implementation of TLS. For the PowerPC ABIs we allocated a
> > > full 4K for the TCB and use negative displacement calculations that work
> > > well with our ISA.
> >
> > I don't see this in glibc. struct pthread seems to be immediately
> > below tcbhead_t, and the latter is not 4k. I'm looking at:
> >
> > https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/powerpc/nptl/tls.h;h=1f3d97a99593afbd3c56318eaa6d7a2d03a59005;hb=HEAD
> >
>
> The key is the following statement from tls.h:
>
> /* The following assumes that TP (R2 or R13) points to the end of the
> TCB + 0x7000 (per the ABI). This implies that TCB address is
> TP - 0x7000. As we define TLS_DTV_AT_TP we can
> assume that the pthread struct is allocated immediately ahead of the
> TCB. This implies that the pthread_descr address is
> TP - (TLS_PRE_TCB_SIZE + 0x7000). */
>
> So struct pthread is allocated immediately ahead of the TCB and grows
> down (to lower addresses) and the TCB alway ends on the byte before R13
> - 0x7000 and grow up (to higher addresses). This is why we always add
> new fields to the front of the TCB struct.
>
> This allow the TCB and struct pthread to grow redundantly from either
> side of R13-0x7000 and allows the TCB field offsets to remain stable
> across releases of the ABI and versions of GLIBC.
>
> The various macros in tls.h handle the details.
The layout as I understand it is not compatible with what you
described; there is certainly no way it can allow growth in both
directions, since one direction grows into the local-exec TLS, which
begins at or just above TP-0x7000.
Here is the layout of TLS, from lowest address to highest address:
1. struct pthread \ These lines 2 and 3 together make up
2. tcbhead_t / the TLS_PRE_TCB_SIZE in tls.h.
3. Nominal TCB, 0 bytes (TLS_TCB_SIZE in tls.h)
4. Local-exec TLS
TP-0x7000 points to the end of 2, or the beginning/end of 3, or the
beginning of 4 (take your pick since they're all the same).
Fields of tcbhead_t can be accessed as ABI since they have a fixed
offset from TP-0x7000, as long as you only add new fields to the
beginning; doing so "pushes struct pthread down", which is harmless.
However, if you access a newly-added field from code assuming it
exists, but you're running with an old glibc version where it did no
exist, you will actually end up accessing the end of struct pthread.
Rich