On Mon, 26 Oct 2026 14:27:53 -0400 David Marchand <[email protected]> wrote:
> > > Series applies cleanly to 26.11-rc1 and every commit builds with - > > > Dwerror=true. The fixes are now first. 08 cherry-picks cleanly onto > > > v24.11 and v25.11, and 06 applies there once 01-05 are in. > > > > > > v26 Warnings on 12 (now 14) and 20 are addressed. The 24 Warning is only > > > partly fixed. > > > > > > > I see only one warning in this patchset 24/25 - is it possible that you > > drop this patch and apply rest of the series? > > IIUC, next-net does not build in GHA since this series was merged: > https://github.com/DPDK/dpdk/actions/runs/38964568382/job/117187710778 > > CI shows: FAILED: [code=1] drivers/libtmp_rte_bus_dpaa.a.p/bus_dpaa_base_qbman_bman.c.o ccache powerpc64le-linux-gnu-gcc -Idrivers/libtmp_rte_bus_dpaa.a.p -Idrivers -I../drivers -Idrivers/bus/dpaa -I../drivers/bus/dpaa -I../drivers/bus/dpaa/base -I../drivers/bus/dpaa/include -I../drivers/bus/dpaa/base/qbman -Ilib/eal/common -I../lib/eal/common -I. -I.. -Iconfig -I../config -Ilib/eal/include -I../lib/eal/include -Ilib/eal/linux/include -I../lib/eal/linux/include -Ilib/eal/ppc/include -I../lib/eal/ppc/include -I../kernel/linux -Ilib/eal -I../lib/eal -Ilib/kvargs -I../lib/kvargs -Ilib/log -I../lib/log -Ilib/metrics -I../lib/metrics -Ilib/telemetry -I../lib/telemetry -Ilib/argparse -I../lib/argparse -Idrivers/common/dpaax -I../drivers/common/dpaax -I../drivers/common/dpaax/caamflib -Ilib/eventdev -I../lib/eventdev -Ilib/ring -I../lib/ring -Ilib/ethdev -I../lib/ethdev -Ilib/net -I../lib/net -Ilib/mbuf -I../lib/mbuf -Ilib/mempool -I../lib/mempool -Ilib/meter -I../lib/meter -Ilib/hash -I../lib/hash -Ilib/rcu -I../lib/rcu -Ilib/timer -I../lib/timer -Ilib/cryptodev -I../lib/cryptodev -Ilib/dmadev -I../lib/dmadev -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Werror -std=c11 -O2 -g -include rte_config.h -Wvla -Wcast-qual -Wdeprecated -Wformat -Wformat-nonliteral -Wformat-security -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wpointer-arith -Wshadow -Wsign-compare -Wstrict-prototypes -Wundef -Wwrite-strings -Wno-packed-not-aligned -Wno-missing-field-initializers -Wno-psabi -D_GNU_SOURCE -fPIC -mcpu=power8 -mtune=power8 -DALLOW_EXPERIMENTAL_API -DALLOW_INTERNAL_API -Wno-format-truncation -Wno-address-of-packed-member -Wno-cast-qual -Wno-pointer-arith -DRTE_COMPONENT_CLASS=pmd_bus -DRTE_COMPONENT_NAME=dpaa -DRTE_LOG_DEFAULT_LOGTYPE=bus.dpaa -MD -MQ drivers/libtmp_rte_bus_dpaa.a.p/bus_dpaa_base_qbman_bman.c.o -MF drivers/libtmp_rte_bus_dpaa.a.p/bus_dpaa_base_qbman_bman.c.o.d -o drivers/libtmp_rte_bus_dpaa.a.p/bus_dpaa_base_qbman_bman.c.o -c ../drivers/bus/dpaa/base/qbman/bman.c ../drivers/bus/dpaa/base/qbman/bman.c: In function ‘bman_release_fast’: ../drivers/bus/dpaa/base/qbman/bman.c:289:51: error: ‘bm_bufs[0].<U6b48>.be_desc.hi’ may be used uninitialized [-Werror=maybe-uninitialized] 289 | r->bufs[0].be_desc.hi = bm_bufs[0].be_desc.hi; | ~~~~~~~~~~~~~~~~~~^~~ ../drivers/bus/dpaa/base/qbman/bman.c:265:26: note: ‘bm_bufs’ declared here 265 | struct bm_buffer bm_bufs[FSL_BM_BURST_MAX]; | ^~~~~~~ ../drivers/bus/dpaa/base/qbman/bman.c:290:51: error: ‘bm_bufs[0].<U6b48>.be_desc.lo’ may be used uninitialized [-Werror=maybe-uninitialized] 290 | r->bufs[0].be_desc.lo = bm_bufs[0].be_desc.lo; | ~~~~~~~~~~~~~~~~~~^~~ ../drivers/bus/dpaa/base/qbman/bman.c:265:26: note: ‘bm_bufs’ declared here 265 | struct bm_buffer bm_bufs[FSL_BM_BURST_MAX]; | ^~~~~~~ AI analysis says: The loop at line 282 starts at i = 0 and is bounded only by num. Without RTE_LIBRTE_DPAA_HWDEBUG nothing guarantees num >= 1, so GCC sees a num == 0 path where bm_bufs[0] is never written but lines 289-290 still read it. The warning is real, not a codegen artifact; ppc/riscv inlining just exposes it. bman_release() and the main version of this function both write entry 0 unconditionally, so restore that: diff bus/dpaa: fix maybe-uninitialized in bman_release_fast The conversion loop is bounded only by num, which is not checked unless RTE_LIBRTE_DPAA_HWDEBUG is set. GCC on ppc64le and riscv therefore warns that bm_bufs[0] may be read uninitialized when copying the first entry into the RCR. Convert entry 0 unconditionally, as bman_release() does. Fixes: <sha> ("<subject of the next-net commit>") --- a/drivers/bus/dpaa/base/qbman/bman.c +++ b/drivers/bus/dpaa/base/qbman/bman.c @@ -279,7 +279,8 @@ bman_release_fast(struct bman_pool *pool, const uint64_t *bufs, if (unlikely(!r)) return -EBUSY; - for (i = 0; i < num; i++) + bm_buffer_set64_to_be(&bm_bufs[0], bufs[0]); + for (i = 1; i < num; i++) bm_buffer_set64_to_be(&bm_bufs[i], bufs[i]); /* bufs[0] overlaps the verb, which bm_rcr_pvb_commit() writes last
