public inbox for linux-pci@vger.kernel.org
 help / color / mirror / Atom feed
From: Niklas Cassel <cassel@kernel.org>
To: "Jingoo Han" <jingoohan1@gmail.com>,
	"Manivannan Sadhasivam" <manivannan.sadhasivam@linaro.org>,
	"Lorenzo Pieralisi" <lpieralisi@kernel.org>,
	"Krzysztof Wilczyński" <kw@linux.com>,
	"Rob Herring" <robh@kernel.org>,
	"Bjorn Helgaas" <bhelgaas@google.com>,
	"Damien Le Moal" <dlemoal@kernel.org>,
	"Frank Li" <Frank.Li@nxp.com>,
	"Niklas Cassel" <cassel@kernel.org>
Cc: "Krzysztof Wilczyński" <kwilczynski@kernel.org>,
	linux-pci@vger.kernel.org
Subject: [PATCH] PCI: dwc: ep: Do not map more memory than needed to raise a MSI/MSI-X IRQ
Date: Mon,  4 Nov 2024 21:51:44 +0100	[thread overview]
Message-ID: <20241104205144.409236-2-cassel@kernel.org> (raw)

In dw_pcie_ep_init() we allocate memory from the EPC address space that we
will later use to raise a MSI/MSI-X IRQ. This memory is only freed in
dw_pcie_ep_deinit().

Performing this allocation in dw_pcie_ep_init() is to ensure that we will
not fail to allocate memory from the EPC address space when trying to raise
a MSI/MSI-X IRQ.

We still map/unmap this memory every time we raise an IRQ, in order to not
constantly occupy an iATU, especially for controllers with few iATUs.
(So we can still fail to raise an MSI/MSI-X IRQ if all iATUs are occupied.)

When allocating this memory in dw_pcie_ep_init(), we allocate
epc->mem->window.page_size memory, which is the smallest unit that we can
allocate from the EPC address space.

However, when writing/sending the msg data, which is only 2 bytes for MSI,
4 bytes for MSI-X, in either case a single writel() is sufficient. Thus,
the size that we need to map is a single PCI DWORD (4 bytes).

This is the size that we should send in to the pci_epc_ops::align_addr()
API. It is align_addr()'s responsibility to return a size that is aligned
to the EPC page size, for platforms that need a special alignment.

Modify the align_addr() call to send in the proper size that we need to
map.

Before this patch on a system with a EPC page size 64k, we would
incorrectly map 128k (which is larger than our allocation) instead of 64k.

After this patch, we will correctly map 64k (a single page). (We should
never need to map more than a page to write a single DWORD.)

Fixes: f68da9a67173 ("PCI: dwc: ep: Use align addr function for dw_pcie_ep_raise_{msi,msix}_irq()")
Signed-off-by: Niklas Cassel <cassel@kernel.org>
---
Feel free to squash this with the patch that it fixes, if you so prefer.

 drivers/pci/controller/dwc/pcie-designware-ep.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
index 9bafa62bed1dc..507e40bd18c8f 100644
--- a/drivers/pci/controller/dwc/pcie-designware-ep.c
+++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
@@ -503,7 +503,7 @@ int dw_pcie_ep_raise_msi_irq(struct dw_pcie_ep *ep, u8 func_no,
 	u32 msg_addr_lower, msg_addr_upper, reg;
 	struct dw_pcie_ep_func *ep_func;
 	struct pci_epc *epc = ep->epc;
-	size_t msi_mem_size = epc->mem->window.page_size;
+	size_t map_size = sizeof(u32);
 	size_t offset;
 	u16 msg_ctrl, msg_data;
 	bool has_upper;
@@ -532,9 +532,9 @@ int dw_pcie_ep_raise_msi_irq(struct dw_pcie_ep *ep, u8 func_no,
 	}
 	msg_addr = ((u64)msg_addr_upper) << 32 | msg_addr_lower;
 
-	msg_addr = dw_pcie_ep_align_addr(epc, msg_addr, &msi_mem_size, &offset);
+	msg_addr = dw_pcie_ep_align_addr(epc, msg_addr, &map_size, &offset);
 	ret = dw_pcie_ep_map_addr(epc, func_no, 0, ep->msi_mem_phys, msg_addr,
-				  msi_mem_size);
+				  map_size);
 	if (ret)
 		return ret;
 
@@ -589,7 +589,7 @@ int dw_pcie_ep_raise_msix_irq(struct dw_pcie_ep *ep, u8 func_no,
 	struct pci_epf_msix_tbl *msix_tbl;
 	struct dw_pcie_ep_func *ep_func;
 	struct pci_epc *epc = ep->epc;
-	size_t msi_mem_size = epc->mem->window.page_size;
+	size_t map_size = sizeof(u32);
 	size_t offset;
 	u32 reg, msg_data, vec_ctrl;
 	u32 tbl_offset;
@@ -616,9 +616,9 @@ int dw_pcie_ep_raise_msix_irq(struct dw_pcie_ep *ep, u8 func_no,
 		return -EPERM;
 	}
 
-	msg_addr = dw_pcie_ep_align_addr(epc, msg_addr, &msi_mem_size, &offset);
+	msg_addr = dw_pcie_ep_align_addr(epc, msg_addr, &map_size, &offset);
 	ret = dw_pcie_ep_map_addr(epc, func_no, 0, ep->msi_mem_phys, msg_addr,
-				  epc->mem->window.page_size);
+				  map_size);
 	if (ret)
 		return ret;
 
-- 
2.47.0


             reply	other threads:[~2024-11-04 20:51 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-11-04 20:51 Niklas Cassel [this message]
2024-11-04 21:13 ` [PATCH] PCI: dwc: ep: Do not map more memory than needed to raise a MSI/MSI-X IRQ Krzysztof Wilczyński
2024-11-04 22:04   ` Niklas Cassel
2024-11-04 22:38     ` Krzysztof Wilczyński
2024-11-15  7:19 ` Manivannan Sadhasivam
2024-11-16 18:37   ` Krzysztof Wilczyński

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20241104205144.409236-2-cassel@kernel.org \
    --to=cassel@kernel.org \
    --cc=Frank.Li@nxp.com \
    --cc=bhelgaas@google.com \
    --cc=dlemoal@kernel.org \
    --cc=jingoohan1@gmail.com \
    --cc=kw@linux.com \
    --cc=kwilczynski@kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=lpieralisi@kernel.org \
    --cc=manivannan.sadhasivam@linaro.org \
    --cc=robh@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox