Message ID | 20240730-tzmem-efivars-fix-v1-1-bb78884810b5@linaro.org (mailing list archive) |
---|---|
State | Superseded |
Headers | show |
Series | firmware: qcom: fix an efivars regression in qseecom | expand |
On Tue, Jul 30, 2024 at 05:38:48PM GMT, Bartosz Golaszewski wrote: > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> > > We currently only correctly convert the virtual address passed by the > caller to qcom_tzmem_to_phys() if it corresponds to the base address of > the chunk. If the user wants to convert some pointer at an offset > relative to that base address, we'll return 0. Let's change the > implementation of qcom_tzmem_to_phys(): iterate over the chunks and try > to call gen_pool_virt_to_phys() just-in-time instead of trying to call > it only once when creating the chunk. > > Fixes: 84f5a7b67b61 ("firmware: qcom: add a dedicated TrustZone buffer allocator") > Reported-by: Johan Hovold <johan+linaro@kernel.org> > Closes: https://lore.kernel.org/lkml/20240729095542.21097-1-johan+linaro@kernel.org/ > Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> > --- > drivers/firmware/qcom/qcom_tzmem.c | 21 +++++++++++++++------ > 1 file changed, 15 insertions(+), 6 deletions(-) > > diff --git a/drivers/firmware/qcom/qcom_tzmem.c b/drivers/firmware/qcom/qcom_tzmem.c > index 17948cfc82e7..1294233a4785 100644 > --- a/drivers/firmware/qcom/qcom_tzmem.c > +++ b/drivers/firmware/qcom/qcom_tzmem.c > @@ -40,7 +40,6 @@ struct qcom_tzmem_pool { > }; > > struct qcom_tzmem_chunk { > - phys_addr_t paddr; > size_t size; > struct qcom_tzmem_pool *owner; > }; > @@ -385,7 +384,6 @@ void *qcom_tzmem_alloc(struct qcom_tzmem_pool *pool, size_t size, gfp_t gfp) > return NULL; > } > > - chunk->paddr = gen_pool_virt_to_phys(pool->genpool, vaddr); > chunk->size = size; > chunk->owner = pool; > > @@ -442,14 +440,25 @@ EXPORT_SYMBOL_GPL(qcom_tzmem_free); > phys_addr_t qcom_tzmem_to_phys(void *vaddr) We should update the kerneldoc here, it currently says: /** * qcom_tzmem_to_phys() - Map the virtual address of a TZ buffer to physical. * @vaddr: Virtual address of the buffer allocated from a TZ memory pool. * * Can be used in any context. The address must have been returned by a call * to qcom_tzmem_alloc(). * * Returns: Physical address of the buffer. */ phys_addr_t qcom_tzmem_to_phys(void *vaddr) "The address must have been returned by a call to qcom_tzmem_alloc()" is no longer a strict requirement, it can be within the buffer now. Otherwise, I believe this should fix the problem you and Johan highlighted and looks ok to me, thanks for the fix. Acked-by: Andrew Halaney <ahalaney@redhat.com>
diff --git a/drivers/firmware/qcom/qcom_tzmem.c b/drivers/firmware/qcom/qcom_tzmem.c index 17948cfc82e7..1294233a4785 100644 --- a/drivers/firmware/qcom/qcom_tzmem.c +++ b/drivers/firmware/qcom/qcom_tzmem.c @@ -40,7 +40,6 @@ struct qcom_tzmem_pool { }; struct qcom_tzmem_chunk { - phys_addr_t paddr; size_t size; struct qcom_tzmem_pool *owner; }; @@ -385,7 +384,6 @@ void *qcom_tzmem_alloc(struct qcom_tzmem_pool *pool, size_t size, gfp_t gfp) return NULL; } - chunk->paddr = gen_pool_virt_to_phys(pool->genpool, vaddr); chunk->size = size; chunk->owner = pool; @@ -442,14 +440,25 @@ EXPORT_SYMBOL_GPL(qcom_tzmem_free); phys_addr_t qcom_tzmem_to_phys(void *vaddr) { struct qcom_tzmem_chunk *chunk; + struct radix_tree_iter iter; + void __rcu **slot; + phys_addr_t ret; guard(spinlock_irqsave)(&qcom_tzmem_chunks_lock); - chunk = radix_tree_lookup(&qcom_tzmem_chunks, (unsigned long)vaddr); - if (!chunk) - return 0; + radix_tree_for_each_slot(slot, &qcom_tzmem_chunks, &iter, 0) { + chunk = radix_tree_deref_slot_protected(slot, + &qcom_tzmem_chunks_lock); - return chunk->paddr; + ret = gen_pool_virt_to_phys(chunk->owner->genpool, + (unsigned long)vaddr); + if (ret == -1) + continue; + + return ret; + } + + return 0; } EXPORT_SYMBOL_GPL(qcom_tzmem_to_phys);